Network Disconnection, Not Lesion Volume, Drives MS Cognition

Long-held assumptions that total lesion burden on MRI is the principal determinant of cognitive decline in patients with multiple sclerosis (MS) were challenged in a recent study published in the Journal of Neuroimaging.1 Researchers demonstrated that lesion-induced network disconnection is a much more powerful predictor of cognitive outcomes in MS than lesion burden alone, addressing the long-standing clinico-radiological paradox in MS, where conventional MRI measures often fail to fully explain the broad spectrum of cognitive difficulties experienced by patients with MS.

Cognitive impairment impacts 40% to 70% of people, even in the early stages of the disease, significantly impairing daily functioning, employability, and overall quality of life.2 The researchers applied high-resolution MRI and computational modeling to generate “disconnectome maps” to provide probabilistic reconstructions of white matter pathways disrupted by lesions.1 This approach extends beyond simply measuring the volume of lesions to assess the functional consequences of lesion location within the brain's complex networks, providing a more comprehensive understanding of how structural damage translates to cognitive dysfunction.

Conventional MRI assessments, focused solely on lesion count or volume, often miss much of the subtle yet consequential network disruption that underlies cognitive decline in MS. | Image credit: jolygon - stock.adobe.com

The cross-sectional study included 30 adults with MS (mean age 37 years, 63% female; 92% relapsing-remitting subtype). Participants underwent high-resolution MRI and cognitive testing with the Brief International Cognitive Assessment for MS (BICAMS), which evaluates processing speed (Symbol Digit Modalities Test [SDMT]), verbal memory (California Verbal Learning Test-II [CVLT-II]), and visuospatial memory (Brief Visuospatial Memory Test–Revised [BVMT-R]).

The study found that even among this relatively mildly disabled cohort, cognitive impairment was widespread, with 42.1% demonstrating deficits in at least one cognitive domain, 21.1% in 2 domains, and 13.2% in all 3 domains. Visuospatial memory was most frequently affected, with 38.7% showing moderate and 30.0% severe deficits. Processing speed and verbal memory impairments were also common, with moderate deficits present in 33.3% and 26.7% and severe deficits in 13.3% and 23.3%, respectively.

The research employed disconnectome mapping, a technique that uses normative tractography data to estimate the likelihood of structural brain disconnection resulting from white matter lesions. This approach enables quantification not just of direct tissue damage, but of the secondary, network-level disruptions caused by these lesions. The study found that disconnectome volume was, on average, 2.3 times larger than the visible lesion volume (median, 36 mL vs 9.4 mL) and was only partially correlated with lesion burden (r = 0.85, P < .001) but was more strongly associated with cognition. In multivariable regression models controlling for age, sex, education, and WML volume, disconnectome volume remained a significant and independent predictor of cognitive impairment (β = –0.41, P = .004), whereas lesion volume lost statistical significance once disconnection was accounted for (P = .28).

It was also found that distinct white matter tracts were related to domain-specific deficits. Poorer processing speed (SDMT) was linked to disconnection in the left posterior cingulum (β = –0.310), right frontal aslant tract (β = –0.175), and left arcuate fasciculus (β = –0.158), pathways supporting attention and executive coordination. Verbal memory (CVLT-II) impairment correlated with damage to the posterior segment of the left arcuate fasciculus (β = –0.232; 95% CI, -0.435 to -0.030), consistent with its role in the phonological loop. Visuospatial memory deficits were notably distributed, with poorer BVMT-R performance linked to disconnection in the right cingulum posterior (β = –0.218, 95% CI, -0.383 to -0.054), left fronto-insular, and other fronto-striatal pathways.

Voxel-wise mapping reinforced and extended these tract-based findings. Slower processing speed was linked to diffuse disconnection across the corpus callosum (overlap 31.65%), the right anterior cingulum, and parts of the superior longitudinal fasciculus. Verbal memory deficits had their strongest anatomical correlates in the corpus callosum (41.42% overlap) and left cingulum anterior, with significant involvement of the primary auditory cortex and default mode network. The most critical anatomical substrate for visuospatial memory impairment was the right superior longitudinal fasciculus III, explaining nearly 20% of the significant disconnection voxel overlap, with involvement of right-lateralized frontoparietal networks as well.

Importantly, patients with multi-domain cognitive impairment also exhibited significantly reduced cortical and subcortical gray matter volumes, suggesting that structural disconnection and regional atrophy may act synergistically to accelerate cognitive decline. These findings align with the "network collapse" theory of cognitive impairment in MS, which proposes that accumulating disconnection leads to loss of network efficiency and subsequent cognitive deterioration.3

Conventional MRI assessments, focused solely on lesion count or volume, often miss much of the subtle yet consequential network disruption that underlies cognitive decline in MS. The researchers present the question: What value does disconnectome mapping offer beyond standard cognitive testing?1 They go on to explain “We propose that the two are complementary, not mutually exclusive. While cognitive testing provides an essential measure of a patient's current functional status (the "what" of impairment), disconnectome analysis offers a potential window into the underlying structural mechanism (the "how")."

Yet, the researchers outline the translational gap that remains for disconnectome mapping, noting that current research-grade software is not yet feasible for widespread clinical application and that further automation and integration into radiological workflows will be needed. However, the findings indicate that quantitative disconnectome metrics could ultimately "help identify patients at higher risk for certain cognitive issues that patients might not spontaneously report, prompting targeted screening and appropriate interventions, and ultimately improving quality of life for those living with MS."

References

1. Dos Santos Silva J, Vasconcelos CCF, Spedo CT, et al. Lesion-based disconnectome explains cognitive outcomes in multiple sclerosis. J Neuroimaging. 2025;35(5):e70081. doi:10.1111/jon.70081
2. McNicholas N, O'Connell K, Yap SM, Killeen RP, Hutchinson M, McGuigan C. Cognitive dysfunction in early multiple sclerosis: a review. QJM. 2018;111(6):359-364. doi:10.1093/qjmed/hcx070
3. Schoonheim MM, Broeders TAA, Geurts JJG. The network collapse in multiple sclerosis: an overview of novel concepts to address disease dynamics. Neuroimage Clin. 2022;35:103108. doi:10.1016/j.nicl.2022.103108