EmulatRx: A Multi-Agent System for Intelligent Clinical Trial Design

Background & Unmet Need

• Randomized controlled trials (RCTs) remain the gold standard for evaluating the efficacy and safety of medical interventions
• However, the costs, logistical complexities, and ethical considerations of conducting a full RCT have led researchers to seek alternative methods for generating robust causal evidence
• Trial emulation, wherein a target clinical trial is simulated as closely as possible using real world data (RWD), provides an alternative method for producing causal evidence and streamlining clinical trial planning
• While trial emulation can generate causal data, each step in the pipeline requires careful domain expertise and rigorous quality checks to ensure valid results
• Unmet Need: Methods for streamlining trial emulation to optimize clinical trial design serve as an alternative for generating robust causal evidence

Technology Overview

• The Technology: EmulatRx is a multiagent system which allows insights to be extracted from RWD to perform target trial emulation
• EmulatRx is comprised of distinct computational nodes—a “Supervisor,” “Trialist,” “Clinician,” “Informatician,” and “Statistician”—each handling specialized tasks
• EmulatRx’s multiagent design can address common obstacles in using RWD, such as missing data, sample size limitations, and covariate imbalances
• PoC Data: In a PoC study, EmulatRx emulated a historical trial evaluating the effectiveness of corticosteroids in managing sepsis within the ICU using the MIMIC-IV database
• EmulatRx identified and relaxed eligibility criteria to maximize the sample size, addressed missing key variables, and refined potential covariates
• The average treatment effect and hazard ratio identified were consistent with the actual RCT

Technology Applications

• Post-market studies for real world effectiveness
• Identification of label expansion opportunities
• Clinical trial planning and design
• Adaptive trial evaluation and design
• Drug repurposing hypothesis generation
• Post-market pharmacovigilance studies

Technology Advantages

• Adheres to standard frameworks to ensure compatibility with heterogenous data sources
• Provides a log of results to enable detailed auditing and reproducibility of results
• Distributed processing enables high scalability
• Integrates standard APIs and GPUs for reduced computing requirements