Background: In knee osteoarthritis, radiographic joint space width (JSW) is frequently used as a surrogate marker for cartilage thickness; however, longitudinal changes in radiographic JSW have shown poor correlations with those of magnetic resonance imaging (MRI) cartilage thickness. There are fundamental differences between the techniques: radiographic JSW represents two-dimensional (2D), weight-bearing, bone-to-bone distance, while on MRI three-dimensional (3D) non-weight-bearing cartilage thickness is measured. In this exploratory study, computed tomography (CT) was included as a third technique, as it can measure bone-to-bone under non-weight-bearing conditions. The objective was to use CT to compare the impact of weight-bearing versus non-weight-bearing, as well as bone-to-bone JSW versus actual cartilage thickness, in the knee. Methods: Osteoarthritis patients ( n = 20) who were treated with knee joint distraction were included. Weight-bearing radiographs, non-weight-bearing MRIs and CTs were acquired before and 2 years after treatment. The mean radiographic JSW and cartilage thickness of the most affected compartment were measured. From CT, the 3D median JSW was calculated and a 2D projectional image was rendered, positioned similarly and measured identically to the radiograph. Pearson correlations between the techniques were derived, both cross-sectionally and longitudinally. Results: Fourteen patients could be analyzed. Cross-sectionally, all comparisons showed moderate to strong significant correlations (R = 0.43–0.81; all p < 0.05). Longitudinal changes over time were small; only the correlations between 2D CT and 3D CT (R = 0.65; p = 0.01) and 3D CT and MRI (R = 0.62; p = 0.02) were statistically significant. Conclusion: The poor correlation between changes in radiographic JSW and MRI cartilage thickness appears primarily to result from the difference in weight-bearing, and less so from measuring bone-to-bone distance versus cartilage thickness.