Predicting MRI Changes in Patients with MS in the Absence of Acute Inflammation
Carmen Tur1, Zhaonan Sun2, Daniel Bradley2, Cynthia Grossman2, Xavier Montalban1, Alex Rovira1, Mar Tintore1, Nolan Campbell2, Elizabeth Fisher2
1Multiple Sclerosis Centre of Catalonia (CEMCAT), Vall d'Hebron, 2Biogen Inc.
Objective:
To predict changes in MRI variables of interest occurring in absence of acute inflammation (AAI), defined as periods with no relapses nor new enlarging T2 lesions.
Background:
Treatments targeting MS disease processes in AAI are lacking, mainly due to the lack of MRI biomarker endpoints suitable for short-term trials.
Design/Methods:
Patients in MS PATHS with ≥2 standardized MRIs over ≥180 days without change in treatment and AAI over follow-up were included. Data were split into training (70%) and testing (30%) sets. Regression models for annualized changes in brain parenchymal fraction (BPF), T1 lesion volume (T1LV), and T1 lesion volume as a proportion of chronic T2 lesion volume (T1LP) were generated in the training set using baseline Patient Determined Disease Steps, prior treatments, treatment group, prior relapse rate, BPF, T1LV, T1LP, cortical and total gray matter and thalamic volume fractions, T2 lesion volume (T2LV), mean normalized T1 intensity within chronic T2 lesions, and spinal cord cross-sectional area (SCCA) as covariates. Model performance was evaluated in the test set via R-squared between predicted and actual change.
Results:
N=1078 patients with mean MRI follow-up of 1.4 years. There were no significant predictors of change in BPF and T1LP (R-squared <0). R-squared and adjusted R-squared for change in T1LV models on the test set were 0.18 and 0.16, respectively; significant covariates were sex (male; P<0.001), T1LV (P<0.001), T2LV (P<0.001), thalamic fraction (P<0.05), and SCCA (P<0.05).
Conclusions:
Baseline demographics, clinical characteristics, and MRI markers did not predict changes in proposed MRI measures of progression in AAI, with the exception of a small proportion of the variance in T1LV change. This finding highlights the need to better understand the processes driving disease progression in AAI. Ability to reliably predict subsequent changes in MRI measures might enable population enrichment in future drug trials targeting disease progression in AAI.