To prospectively determine the reproducibility of quantitative magnetic resonance (MR) imaging biomarkers of the morphology and composition (spin lattice relaxation time in rotating frame [T1-ρ], T2) of knee cartilage in a multicenter multivendor trial involving patients with osteoarthritis (OA) and asymptomatic control subjects.
Materials and Methods
This study was HIPAA compliant and approved by the institutional review committees of the participating sites, with written informed consent obtained from all participants. Fifty subjects from five sites who were deemed to have normal knee joints (n = 18), mild OA (n = 16), or moderate OA (n = 16) on the basis of Kellgren-Lawrence scores were enrolled. Each participant underwent four sequential 3-T knee MR imaging examinations with use of the same imager and with 2–63 days (median, 18 days) separating the first and last examinations. Water-excited three-dimensional T1-weighted gradient-echo imaging, T1-ρ imaging, and T2 mapping of cartilage in the axial and coronal planes were performed. Biomarker reproducibility was determined by using intraclass correlation coefficients (ICCs) and root-mean-square coefficients of variation (RMS CVs, expressed as percentages).
Morphometric biomarkers had high reproducibility, with ICCs of 0.989 or greater and RMS CVs lower than 4%. The largest differences between the healthy subjects and the patients with radiographically detected knee OA were those in T1-ρ values, but precision errors were relatively large. Reproducibility of T1-ρ values was higher in the thicker patellar cartilage (ICC range, 0.86–0.93; RMS CV range, 14%–18%) than in the femorotibial joints (ICC range, 0.20–0.84; RMS CV range, 7%–19%). Good to high reproducibility of T2 was observed, with ICCs ranging from 0.61 to 0.98 and RMS CVs ranging from 4% to 14%.
MR imaging measurements of cartilage morphology, T2, and patellar T1-ρ demonstrated moderate to excellent reproducibility in a clinical trial network.