We investigated whether the baseline level and overtime changes of white matter hyperintensities (WMH) would be associated with cognitive decline over three years in non-demented older adults with memory complaints. 109 participants with baseline magnetic resonance imaging (MRI) and follow-up cognitive assessments up to 3-year were included; among them, 82 also had a follow-up MRI assessment over three years. WMH volume was obtained by an automated segmentation algorithm. Baseline WMH volumes and change between baseline and follow-up WMH were related to cognitive scores over time using mixed-effect linear regressions. Secondary stratified analyses according to Clinical Dementia Rating (CDR) status, APOE4 status, and presence of amyloid in the brain were conducted using similar regression models. Change in WMH volume overtime was associated with declines in COWAT (β=-0.239; 95% CI=-0.381, -0.096, p=0.001). Baseline WMH was not associated to any of the cognitive tests. Secondary analysis found that baseline WMH was associated to declines in TMT-A in APOE4 non-carriers (β=0.343; 95% CI=0.121, 0.564, p=0.003) and CDR 0 groups (β=0.307; 95% CI=0.095, 0.519, p=0.005); in CDR 0 group, overtime changes in WMH was associated to declines on both TMT-A (β=0.698; 95% CI=0.270, 1.126, p=0.002) and TMT-B (β=2.573; 95% CI=1.200, 3.947, p<0.001). Changes in WMH volume are associated with declines in information processing speed and executive function in non-demented older adults with memory complaints. Overtime changes in WMH volume is probably a better determinant of cognitive function in the elderly than baseline WMH volume.