Abstract Objectives Knee joint distraction (KJD) has been shown to result in long-term clinical improvement and short-term cartilage restoration in young OA patients. The objective of the current study was to evaluate MRI cartilage thickness up to 10 years after KJD treatment, using a 3D surface-based approach. Methods Twenty end-stage knee OA patients were treated with KJD. MRI scans (1.5 T) were performed before and at 1, 2, 5, 7, and 10 years after treatment. Tibia and femur cartilage segmentation and registration to a canonical surface were performed semi-automatically. Statistical parametric mapping with linear mixed models was used to analyse whole-joint changes. The influence of baseline patient characteristics was analysed with statistical parametric mapping using linear regression. Relevant weight-bearing parts of the femur were selected to obtain the average cartilage thickness in the femur and tibia of the most- (MAC) and least-affected compartment. These compartmental changes over time were analysed using repeated measures ANOVA; missing data was imputed. In all cases, P <0.05 was considered statistically significant. Results One and 2 years post-treatment, cartilage in the MAC weight-bearing region was significantly thicker than pre-treatment, gradually thinning after 5 years, but still increased at 10 years post-treatment. Long-term results showed that areas in the least-affected compartment were significantly thicker than pre-treatment. Male sex and more severe OA at baseline somewhat predicted shorter-term benefit (P >0.05). Compartmental analyses showed significant short- and long-term thickness increase in the tibia and femur MAC (all P <0.05). Conclusion KJD results in significant short- and long-term cartilage regeneration, up to 10 years post-treatment. Trial registration Netherlands Trial Register, https://www.trialregister.nl, NL419.