Case studies
From research to practice: AI innovations implemented and validated in real clinical trial settings.
Improving phase 2 Trial Precision in Parkinson's Disease Using Digital Twin-Derived Prognostic Covariates
Joint development with Sanofi
Context
Qairnel and Sanofi partnered to evaluate how digital twin technology could improve trial precision in Phase 2 Parkinson’s disease (PD) trials.
The work used data from the MOVES-PD randomized controlled trial, which investigated venglustat in patients with GBA1-associated Parkinson’s disease.
PD are characterized by highly variable progression across patients, making it difficult to detect treatment effects — particularly in mid-stage (Phase 2) studies with limited sample sizes.
Challenge
The central challenge was clinical variability:
- Trial participants progress at very different rates
- Standard statistical models lack explanatory variables to control for this heterogeneity
- Trial outcomes with large variance are hard to interpret
- Confidence in efficacy signals is reduced
- Larger sample sizes are often required to compensate
For sponsors, this creates a major issue: uncertainty in decision-making at a critical stage of development.
Approach
Qairnel applied CourseMap AI its patented digital-twin technology, to analyze trial data:
- Built predictive digital twins to estimate individual disease trajectories from baseline data of each participant
- Generated patient-specific prognostic covariates from these predicted trajectories
- Integrated these covariates into standard statistical models (MMRM)
- Compared multiple modeling approaches (no covariates vs. standard vs. digital-twin enhanced)
This approach effectively explains part of the variability upfront, allowing statistical models to isolate treatment effects more clearly.
Outcomes
The inclusion of digital twin–derived covariates delivered clear, quantifiable improvements:
Improved Precision
- Narrower confidence intervals of the treatment effect, reaching statistical significance
- Better interpretability of the trial outcome
- Clearer confirmation of a negative treatment effect vs placebo
Decreased Effective Sample Size
- ~15–16% gain in statistical power
- Compared to ~5% using standard covariates
- Potential reduction in Phase 3 sample size from 361 → 304 patients
Scientific and Industry Validation
- Co-developed and co-published with Sanofi
- Presented at the ADPD 2025 conference
- Aligned with regulatory guidance on prognostic covariate adjustment