Background: Change in knee cartilage volume is frequently used as a proxy for change in knee joint space width over time, but longitudinal data on these associations is limited. We aimed to determine whether change in knee cartilage volume, new or worsening meniscal extrusion (ME), meniscal tears and cartilage defects over 2.4 years correlated with change in joint space width (JSW) over 5 years in older community dwelling adults. Methods: Participants (n = 153) had their right knee imaged using MR imaging and x-ray at baseline, and after 2.4 years (MRI) and 5 years (x-ray). Cartilage volume, cartilage defects, meniscal extrusions and meniscal tears were assessed on sagittal T1-weighted fat-suppressed MRI. JSW was assessed using standard fixed semi-flexed view radiographs, and scored on those with adequate alignment. Results: Participants were 51-79 (mean 62) years old; 48 % were female. Cartilage volume reduced over time (medial -134 ± 202 μL/year, lateral -106 ± 165 μL/year, p < 0.001), as did JSW (medial -0.05 ± 0.16 mm/year, lateral -0.12 ± 0.24 mm/year, p < 0.001). In multivariable analysis, the only consistent predictor of change in JSW was new or worsening ME (medial tibia R2 3.1 %, p = 0.031; medial femur R2 3.2 %, p = 0.024); change in cartilage volume correlated with change in JSW laterally (R2 4.8 %, p = 0.007) and was borderline medially (R2 2.2 %, p = 0.064); there was no association for meniscal tears or cartilage defects. The magnitude of these associations were similar albeit somewhat greater for ME in participants with radiographic OA (R2 6.2 %, p = 0.017). Conclusion: Change in ME and cartilage volume weakly predict change in JSW, but the vast majority of the variation remains unexplained. Since MRI examines cartilage directly while radiographs examine it indirectly, these results cast doubt on the validity of using JSW as a proxy measure of cartilage loss.