Background/Purpose: Imaging biomarkers that predict relevant clinical endpoints, such as knee replacement (KR), are valuable tools for knee osteoarthritis prognosis. Currently, measurement of minimum radiographic joint space width (mJSW) is the accepted standard for testing the efficacy of disease modifying drugs (DMOADs) in clinical trials. The purpose of this study was to compare the predictive power of magnetic resonance imaging (MRI)-based longitudinal measures of cartilage loss and mJSW for KR.

Methods: We studied knees from Osteoarthritis Initiative (OAI) participants who received a KR between 24 and 60 month (M) follow-up (confirmed by radiography and/or review of hospital records). A matched control knee that did not receive a KR during this period was selected for each case, with the same sex, similar age (within 5 years), and the same baseline Kellgren Lawrence (KL) grade (central X-ray readings; strata 0-1, 2, 3, 4). Knees were not discordant for medial vs. lateral semi-quantitative joint space narrowing (JSN). Medial femorotibial compartment (MFTC) cartilage thickness change was determined from a sagittal 3 Tesla double echo steady state water excitation (DESSwe) MRI sequence, and, mJSW from fixed flexion radiography. The time points prior to KR (T0) and that 12 months earlier (T-1) were analyzed in case knees (i.e. in a knee with KR between 24M and 36M: 24M=T0;12M=T-1), and the same time points in control knees, with blinding to acquisition order. P-values for differences in longitudinal change between case/control pairs were assessed using the Wilcoxon signed rank test, and the area (AUC) under the receiver operation curve (ROC)

Results: 261 knees of 225 OAI participants received a KR (between 24M and 60M). Of these, 93 had central X-ray readings, JSW, and MRI readings available at both T0 and T-1, and a matched control based on the above criteria (38 men, 55 women, age 63.4±9.3, BMI 30.2±4.9). Medial (MFTC) cartilage loss (over a 12 month period prior to KR), measured with MRI, was significantly greater in KR case than in control knees (-0.13±0.29 vs. -0.04±0.15mm; p=0.038, AUC=0.57). The differences in longitudinal change was borderline significant for the central MFTC (-0.24±0.57 vs. -0.07±0.23 mm; p=0.079, AUC=0.57) and was not significant for the change in radiographic mJSW (-0.19±0.91 vs. 0.01±0.58mm; p=0.397, AUC=0.55). The difference in cartilage loss (MRI) between KR case and control knees was particularly strong in knees with early radiographic OA (KLG2: n=22; MFTC p=0.008; AUC=0.72), whereas no difference in mJSW change was detectable in KLG2 case/control pairs (p=0.126; AUC=0.58).

Conclusion: Medial compartment longitudinal cartilage thickness loss, measured with MRI, predicts knee replacement as a clinical endpoint. In contrast, no significant difference in the 12 month change of radiographic mJSW was detected between case and control knees in the above sample. Given the relationship with an important clinical endpoint (i.e. knee replacement), MRI-based measures of cartilage thickness change are useful imaging biomarkers for clinical trials, for instance to demonstrate the efficacy of a DMOAD. The current findings also support the concept that treatments that slow cartilage loss may delay or prevent KR.