Associations of white matter hyperintensity regression and progression with cognitive and structural brain changes: a population-based cohort study

White matter hyperintensities (WMHs) of presumed vascular origin are common neuroimaging markers of brain aging. While WMHs typically progress over time, WMHs regression is increasingly being reported. However, its clinical relevance remains unclear, particularly in community-based populations. We investigated associations of WMHs regression and progression, including total WMH, periventricular (PWMHs), and deep WMHs (DWMHs), with longitudinal cognitive decline and structural brain atrophy. We analyzed longitudinal data from 2496 UK Biobank participants (median [IQR] age 61 [56–67] years, 52.8% female) without dementia or conditions confounding WMH measures. Brain MRI and cognitive assessments were conducted at baseline and follow-up (2–3 years apart). Absolute WMH volume change was categorized into tertiles representing regression (Tertile 1), stability (Tertile 2), and progression (Tertile 3). Cognitive outcomes included processing speed (reaction time) and fluid intelligence; brain structural outcomes included volumes (% intracranial volume) of total brain, gray matter, white matter, and hippocampi. Associations were examined using multivariable-adjusted linear mixed-effects models, with moderation analyses for age and sex. At baseline, WMH regression and progression groups showed larger WMH volumes than the stable group (p < 0.001). Over the follow-up, total WMH regression (β = 0.029 [95% CI 0.003 to 0.054], p = 0.027) and PWMH regression (β = 0.029 [0.003 to 0.054], p = 0.027) were associated with slower decline in processing speed, but not significantly associated with decline in fluid intelligence or brain structures (all p > 0.05). In contrast, PWMH progression was associated with faster declines in total brain (β=−0.078 [− 0.143 to − 0.012], p = 0.020) and white matter volumes (β=−0.048 [− 0.092 to − 0.005], p = 0.030). Progression in total WMH (β=−0.0016 [− 0.002 to − 0.001], p < 0.001), PWMH (β=−0.0019 [− 0.003 to − 0.001], p < 0.001), and DWMH (β=−0.0008 [− 0.001 to − 0.0002], p = 0.011) was linked to accelerated hippocampal atrophy, with the effects of total WMH and PWMH progression significant only in participants < 65 years (p-interaction < 0.05). Compared with stable WMH, WMH regression was associated with improved processing speed and comparable brain structure, whereas WMH progression was associated with accelerated brain atrophy, particularly hippocampal atrophy. These associations were predominantly driven by periventricular lesions. These findings highlight the dynamic nature and potential reversibility of microvascular injury, supporting WMH regression as a promising intervention target for preserving cognition and brain integrity.