To assess the reproducibility of 3D delayed gadolinium-enhanced MRI of cartilage (dGEMRIC) at 3 T in early stage knee osteoarthritis (OA) patients. In 20 patients, 3D dGEMRIC at 3 T was acquired twice within 7 days. To correct for patient motion during acquisition, all images were rigidly registered in 3D. Eight anatomical cartilage ROIs were analysed on both images of each patient. Capability of dGEMRIC to yield T1 maps that reproducibly distinguish spatial differences in cartilage quality was assessed in two ROIs within a single slice in each patient. Reproducibility was assessed using ICCs and Bland-Altman plots. ICCs ranged from 0.87 to 0.95, indicating good reproducibility. T1 maps revealed reproducible spatial differences in cartilage quality (ICC 0.79). Based on the Bland-Altman plots, we defined a threshold of 95 ms to determine if a change in dGEMRIC outcome in longitudinal research was statistically significant. 3D knee dGEMRIC at 3 T combined with 3D image registration is a highly reproducible measure of cartilage quality in early stage OA. Therefore, dGEMRIC may be a valuable tool in the non-invasive evaluation of cartilage quality changes in longitudinal research in patients with early stage OA and focal cartilage defects. aEuro cent Delayed gadolinium-enhanced magnetic resonance imaging of cartilage (dGEMRIC) can assess osteoarthritis aEuro cent dGEMRIC yields highly reproducible T1 values in early stage osteoarthritic patients aEuro cent A threshold was established to determine significant changes in dGEMRIC outcomes aEuro cent dGEMRIC can be used to evaluate cartilage quality in longitudinal research.