PII-013 - OPTIMIZATION OF PARKINSON'S DISEASE DIGITAL BIOMARKERS THROUGH CLINICAL TRIAL SIMULATIONS.

J. Hogan¹, S. Robey¹, S. Chatterjee¹, J. Ren¹, M. Dockendorf¹, K. Yee²; ¹Merck & Co., Inc., Rahway, NJ, USA, ²Merck & Co., Inc..

Background: Previous Parkinson's Disease (PD) progression modeling efforts have demonstrated that MDS Unified Parkinson's Disease Rating Scales (MDS-UPDRS), an accepted registrational endpoint for PD, provides the most sensitive clinical evaluation of disease progression. However, the high degree of variability requires large and lengthy clinical trials (>300 subjects; 3 years) to detect significant disease modifying (DM) effects1. Digital biomarkers offer an opportunity to evaluate clinical function with more objectivity, precision, and frequency to improve the statistical powering of clinical trials for disease modifying (DM) therapies.

Methods: A trial simulation approach was used to investigate factors that may meaningfully improve power of a digital biomarker relative to MDS-UPDRS Part III. A disease progression model for MDS-UPDRS Part III fit to subject level longitudinal data from the Critical Path for Parkinson’s Integrated Parkinson’s Database was used as a benchmark for the hypothetical digital biomarker, and properties like sensitivity to disease progression rate, inter-individual variability (IIV), and test-retest variability were assessed. The impact of sampling frequency, trial size, and duration on trial design were also evaluated. The power to detect a 25% DM effect was determined with linear mixed effects modeling.

Results: Population mean progression rate and the related standard deviation had the largest impact on power. A digital biomarker capable of detecting disease progression 1.5-fold faster than MDS-UPDRS Part III resulted in a 55.7% reduction in the number of subjects needed to detect a significant slowing of disease progression in a one-year trial. Similarly, a biomarker with 0.5-fold reduced standard deviation on progression rate can support a trial with 70.0% fewer subjects/arm. Other factors explored, including the sampling frequency and test-retest variability, had minimal impact on statistical power in these simulations.

Conclusion: Development of digital biomarkers for the purpose of increased statistical power to detect PD progression and effects of DM treatments should focus on endpoints that can more readily and rapidly measure these changes over time while having greater precision and a differentiated temporal relationship compared to MDS-UPDRS.