Cognitive impairment (CI) affects up to 70% of multiple sclerosis (MS) patients. Although several magnetic resonance imaging (MRI) correlates of CI have been suggested, their relative contribution to explain CI requires further investigation.
To evaluate the combined contribution of white matter (WM) lesions, gray matter (GM) atrophy and resting state (RS) functional (f) MRI abnormalities in explaining CI in a large cohort of MS patients.
Brain 3T dual-echo, 3D T1-weighted and RS fMRI scans were acquired from 100 healthy controls (HC) and 276 MS patients. All MS patients underwent the Rao’s battery. CI was defined by ≥2 tests with a z-score<-1.5. Distribution of brain WM lesions, GM atrophy and RS functional connectivity (FC) abnormalities within the default mode (DMN) and salience (SN) networks were compared between HC and MS patients at a voxel level. Using sex-, age- and phenotype-adjusted stepwise logistic regression models, the role of WM lesions (model 1), GM atrophy (model 2), RS FC (model 3) and their combination (model 4) in explaining CI was investigated. Model performances were assessed by the area under the curve (AUC).
Eighty-three MS patients had CI. In model 1, lesions in left (L) superior longitudinal fasciculus (SLF) (odds ratio [OR]=1.84), L medial lemniscus (OR=1.79) and L inferior longitudinal fasciculus (OR=1.57) predicted CI (p≤0.009). In model 2, L precuneus (OR=0.52) and L caudate (OR=0.56) volumes predicted CI (p≤0.007). In model 3, increased RS FC in L caudate (DMN) (OR=1.77) and decreased RS FC in right (R) thalamus (DMN) (OR=0.66) and L inferior frontal gyrus (IFG) (SN) (OR=0.68) predicted CI (p≤0.02). In model 4, R middle cerebellar peduncle (OR=2.05) and L SLF (OR=1.84) lesions, L precuneus atrophy (OR=0.46), increased RS FC in L caudate (DMN) (OR=1.64), and decreased RS FC in L IFG (SN) (OR=0.64) predicted CI (p≤0.02). Compared to demographic and clinical variables only (AUC=0.73), the separate models performed significantly better (AUC=0.82, 0.81 and 0.80, respectively, p≤0.003), with model 4 having the best performance (AUC=0.86, p<0.001).
The combination of multiparametric MRI techniques contributes to better understand the structural and functional substrates of cognitive dysfunction in MS patients. The accumulation of focal WM lesions and GM atrophy in strategic brain regions together with maladaptive functional mechanisms explains CI in MS.