# Potential role of MRI to optimize clinical trial design for progressive supranuclear palsy and corticobasal degeneration

## TL;DR

MRI can improve clinical trial design for progressive supranuclear palsy and corticobasal degeneration by reducing required sample sizes and providing objective outcome measures.

## Contribution

MRI-based models accurately identify PSP/CBD and reduce trial sample sizes by up to 87% compared to clinical methods.

## Key findings

- MRI signatures cut PSP trial sample size by 50% and CBD by 87%.
- MRI-based models track disease progression more sensitively than clinical scales.
- Shared midbrain/pontine atrophy distinguishes PSP/CBD from other syndromes.

## Abstract

•MRI-based enrichment broadens trial selection for PSP and CBD.•Data-driven MRI signatures provide sensitive and objective outcome measures.•This method cuts PSP sample size by 50 % and CBD by 87 % vs clinical selection/endpoints.

MRI-based enrichment broadens trial selection for PSP and CBD.

Data-driven MRI signatures provide sensitive and objective outcome measures.

This method cuts PSP sample size by 50 % and CBD by 87 % vs clinical selection/endpoints.

Progressive supranuclear palsy (PSP) and corticobasal degeneration (CBD) are 4–repeat tauopathies (4RT) presenting with overlapping syndromes. Imperfect clinicopathological associations increase sample-size demands in clinical trials. We test whether MRI can enrich trials for PSP/CBD and provide sensitive MRI-based outcome measures.

Longitudinal cohort analysis including participants from the 4 Repeat Tauopathy Neuroimaging Initiative (4RTNI) and phase 2/3 Davunetide trial (DAV). An MRI model trained on autopsy-confirmed cases predicted PSP (MRI-PSP) or CBD (MRI-CBD); corticobasal syndrome (CBS) with positive Alzheimer’s biomarkers was reclassified. Clinical scales and MRI-derived thickness/volume were analyzed with linear mixed-effects models. We derived data-driven MRI-signatures (optimal regional combinations) to minimize required trial sample sizes.

206 participants from 4RTNI (n = 106 with Richardson’s syndrome [RS], CBS, or nonfluent/agrammatic primary progressive aphasia [nfvPPA]) and DAV (n = 100 with RS). In 4RTNI, 49 participants were predicted MRI-PSP and 43 MRI-CBD. 76% of MRI-PSP had RS, 24% had CBS/nfvPPA; 66% of MRI-CBD had CBS. PSP and CBD signatures shared midbrain/pontine atrophy but differed in cortical involvement. PSP signature correlated strongly with 12-month change on the PSP Rating Scale (β = –0.59, p < 0.001). MRI-based signatures reduced the estimated sample sizes required to detect 30% reduction in progression over 12-months by 50% for MRI-PSP and 87% for MRI-CBD, compared with clinical outcomes. In DAV, feasibility was replicated.

MRI-derived models can identify PSP or CBD with high accuracy, and MRI-based signatures track progression more sensitively than established clinical outcomes. Incorporating these tools into therapeutic trial design could reduce sample sizes and enable more inclusive disease-modifying trials for 4RT.