Free Water Collapse in Limbic White Matter Predicts Memory Decline in Aging and Alzheimer Disease

Highlight

1) In the largest pooled analysis to date (4,467 participants; 9,208 longitudinal cognitive sessions), diffusion MRI-derived free water (FW) outperformed conventional white matter metrics in predicting baseline cognition and future decline across domains, most strongly memory.

2) Limbic tracts — the cingulum and fornix — showed the strongest FW-cognition associations; FW interacted with APOE ε4, amyloid status, and hippocampal atrophy to accentuate risk for accelerated decline.

3) These results underscore the value of FW correction in diffusion MRI studies and support integrating FW measures with canonical Alzheimer disease (AD) biomarkers in models of neurodegeneration.

Background

Alzheimer disease research has traditionally focused on cortical pathology (amyloid and tau) and gray matter atrophy as proximate causes of cognitive decline. However, accumulating evidence implicates white matter abnormalities in the pathophysiology and clinical expression of aging-related cognitive impairment and AD. Diffusion-weighted MRI (dMRI) provides microstructural metrics of white matter integrity, but extracellular free water (FW) contamination can obscure true tissue-specific changes. Free water imaging—modeling a separate extracellular compartment—aims to disentangle extracellular fluid changes from tissue-restricted diffusion, potentially improving sensitivity to neurodegenerative processes relevant to cognition.

Study design

This prognostic, multisite secondary-data analysis pooled harmonized data from nine cohorts collected between September 2002 and November 2022. Inclusion required participants aged ≥50 years with quality-controlled dMRI, domain-specific cognitive composite z scores, demographics (age, sex, education, race/ethnicity), APOE haplotype, and clinical diagnosis (cognitively unimpaired, mild cognitive impairment [MCI], or AD dementia). The analysis included 4,467 participants and 9,208 longitudinal cognitive assessments. Imaging modalities included dMRI (with FW correction), T1-weighted MRI for gray matter/hippocampal volumes, and amyloid and tau PET for subsets. Primary outcomes were cross-sectional cognitive performance and longitudinal cognitive decline across domains (memory, executive function, language, visuospatial abilities, processing speed).

Key findings

This analysis presents several converging results establishing white matter FW as a key correlate of cognitive impairment and decline in aging and AD.

Population characteristics

Of the 4,467 participants, mean age was 74.3 years (SD 9.2); 60.4% were female. Clinical classification included 3,213 cognitively unimpaired, 972 MCI, and 282 with AD dementia at baseline.

Free water is the strongest white matter predictor of cognition

Across tracts and cognitive domains, FW measures showed the largest effects for both cross-sectional cognition and longitudinal decline compared with conventional diffusion metrics (e.g., fractional anisotropy, mean diffusivity) and uncorrected measures. Effect sizes were particularly pronounced for memory.

Limbic tract vulnerability: cingulum and fornix

Limbic white matter tracts exhibited the strongest associations. Notable effect estimates included:

• Memory performance: cingulum FW β = -0.718 (P < .001); fornix FW β = -1.069 (P < .001).
• Memory decline: cingulum FW β = -0.115 (P < .001); fornix FW β = -0.153 (P < .001).

These values indicate that higher FW in these tracts corresponds to worse baseline memory and faster memory decline. The fornix, a primary efferent tract of the hippocampus, showed particularly large cross-sectional and longitudinal effects.

Interactions with established AD biomarkers and risk factors

FW measures did not act in isolation. Interaction models revealed that FW effects on cognition were modified by clinical diagnosis, gray matter atrophy, APOE ε4 carriage, and amyloid status. Selected interaction findings included:

• Fornix FW × hippocampal volume: β = 10.598 (P < .001), indicating that elevated FW coupled with hippocampal atrophy predicts substantially poorer memory.
• Cingulum FW × SPARE-AD (a multivariate MRI index of AD-like atrophy): β = -0.532 (P < .001), indicating compounding effects of cortical-pattern atrophy and limbic FW on memory performance.
• Inferior temporal gyrus transcallosal tract FW × baseline diagnosis: β = -0.537 (P < .001), showing differential vulnerability by clinical stage.

Moreover, FW effects were accentuated in APOE ε4 carriers and amyloid-positive individuals, suggesting that extracellular water accumulation may interact with molecular AD pathology to amplify cognitive decline.

Domain specificity

Although FW was associated with decline across multiple cognitive domains, memory exhibited the strongest and most consistent relationships with limbic tract FW. Other domains (executive function, processing speed) showed significant but smaller associations and more variable tract localization.

Comparative performance and implications for dMRI processing

The findings emphasize that FW-corrected measures capture distinct and clinically meaningful aspects of white matter pathology that standard diffusion metrics may miss. This argues for routine application of FW-correction pipelines in studies examining cognitive aging and AD.

Expert commentary and interpretation

These results resonate with neuroanatomical and pathophysiological models in which limbic circuit disruption contributes to memory impairment. The fornix and cingulum are anatomically central to hippocampal–cingulo-hippocampal networks; their microstructural compromise plausibly undermines memory encoding and retrieval. The large sample size and multisite harmonization strengthen generalizability and statistical power to detect tract-specific effects and interactions with molecular markers.

Biological plausibility

Free water increases may reflect a blend of pathologies: neuroinflammation with vasogenic edema, degeneration-related enlargement of extracellular space, and interstitial fluid dysregulation related to impaired clearance of metabolites (including amyloid and tau). FW may therefore be a sensitive trans-diagnostic marker of extracellular changes preceding or accompanying tissue loss.

Clinical relevance

From a translational perspective, FW imaging could serve several roles: risk stratification among cognitively unimpaired older adults, a prognostic biomarker for rate of decline in MCI, and a complementary outcome in trials targeting neuroinflammation, glymphatic clearance, or white matter protection. Because FW interacts with amyloid and APOE ε4, combined biomarker models could refine individualized prognoses and trial enrichment strategies.

Limitations and caveats

Important limitations warrant emphasis. First, pooled multisite data improves power but introduces heterogeneity in acquisition, which the authors addressed through harmonization and quality control; residual scanner-related variability may remain. Second, FW is an indirect metric—without histopathology it cannot definitively specify the biological substrate (inflammation vs. atrophy vs. fluid shift). Third, although interaction models implicate synergistic effects with amyloid, cross-sectional PET availability limits causal inference about temporal ordering. Finally, clinical application requires standardization of FW estimation and normative references across platforms.

Conclusions and implications for research and practice

This large-scale multisite study identifies free water in limbic white matter—particularly the cingulum and fornix—as robust predictors of both baseline memory performance and longitudinal decline in aging and AD-related impairment. FW-corrected diffusion metrics capture clinically meaningful extracellular changes that complement gray matter and molecular biomarkers. Future work should prioritize longitudinal multimodal imaging with histopathologic correlation when possible, standardize FW pipelines for clinical research, and evaluate FW as an outcome measure in intervention trials targeting neuroinflammation or protein clearance pathways.

Practical takeaways

Clinicians and researchers should consider FW-corrected diffusion metrics when evaluating white matter contributions to cognitive impairment. Limbic white matter FW may be a sensitive early marker and a prognostic indicator, particularly when combined with hippocampal volume, amyloid/tau PET, and APOE status. Incorporation of FW into multimodal biomarker frameworks could sharpen diagnostic and prognostic precision in aging and AD research.

Funding and trial registration

Details on cohort funding, participating studies, and trial registrations are provided in the original publication (Peter et al., JAMA Neurol. 2025). Readers should consult that article and supplementary materials for full funding acknowledgments and cohort-specific trial identifiers.

Selected references

1. Peter C, Sathe A, Shashikumar N, et al. White Matter Abnormalities and Cognition in Aging and Alzheimer Disease. JAMA Neurol. 2025;82(8):825-836. doi:10.1001/jamaneurol.2025.1601 IF: 21.3 Q1 B1.

2. Pasternak O, Sochen N, Gur Y, Intrator N, Assaf Y. Free water elimination and mapping from diffusion MRI. Magn Reson Med. 2009;62(3):717-730. doi:10.1002/mrm.22055 IF: 3.0 Q2 B3.

3. Bubb EJ, Metzler-Baddeley C, Aggleton JP. The cingulum bundle: Anatomy, function, and dysfunction. Neurosci Biobehav Rev. 2018;92:104-127. doi:10.1016/j.neubiorev.2018.05.008 IF: 7.9 Q1 B1.

4. Jack CR Jr, Bennett DA, Blennow K, et al. NIA-AA Research Framework: Toward a biological definition of Alzheimer’s disease. Alzheimers Dement. 2018;14(4):535-562. doi:10.1016/j.jalz.2018.02.018 IF: 11.1 Q1 B1.