It has been suggested that white matter hyperintensity lesions (WMHs), which typically progress over time, can also regress, and that this might be associated with favorable cognitive performance. We determined the prevalence of WMH regression in patients with cerebral small vessel disease (SVD) and examined which demographic, clinical, and radiological markers were associated with this regression. We used semi-automated lesion marking methods to quantify WMH volume at multiple timepoints in three cohorts with symptomatic SVD; two with moderate-to-severe symptomatic SVD (the SCANS observational cohort and the control arm of the PRESERVE interventional trial) and one with mild-to-moderate SVD (the RUN DMC observational cohort). Mixed-effects ordered logistic regression models were used to test which factors predicted participants to show WMH regression. No participants (0/98) in SCANS, 6/42 (14.3%) participants in PRESERVE, and 6/276 (2.2%) in RUN DMC showed WMH regression. On multivariate analysis, only lower WMH volume (OR: 0.36, 95% CI: 0.23-0.56) and better white matter microstructural integrity assessed by fractional anisotropy using diffusion tensor imaging (OR: 1.55, 95% CI: 1.07-2.24) predicted participant classification as regressor versus stable or progressor. Only a small proportion of participants demonstrated WMH regression across the three cohorts, when a blinded standardized assessment method was used. Subjects who showed regression had less severe imaging markers of disease at baseline. Our results show that lesion regression is uncommon in SVD and unlikely to be a major factor affecting the use of WMH quantification as an outcome for clinical trials.