Total Knee Arthroplasty: Diagnostic Accuracy of Patterns of Synovitis at MR Imaging.

Purpose To determine the diagnostic accuracy of magnetic resonance (MR) imaging for differentiating synovial patterns in patients with total knee arthroplasty (TKA), whether diagnostic accuracy differs in index versus revision TKA, and interobserver and intraobserver reliability for assessment of synovial patterns at MR imaging. Materials and Methods This retrospective cross-sectional study included 108 consecutive patients with TKAs who underwent MR imaging within 1 year prior to revision surgery from 2012 to 2014. Institutional review board approval was obtained, with waiver of the need to obtain informed consent. MR images were reviewed, and cases were qualitatively categorized by the appearance of the synovium as one of the following: frondlike and hypertrophied (particle-induced synovitis), lamellated and hyperintense (infection), and a homogeneous effusion with the signal intensity of fluid (nonspecific synovitis). The MR imaging appearance was compared with surgical and microbiology reports as the reference standard to determine the sensitivity, specificity, and positive and negative predictive values for the index TKA and revision TKA cohorts. Results For all patients combined, MR imaging had 0.907-0.930 sensitivity and 0.723-0.738 specificity for a surgical diagnosis of complications related to polyethylene wear (including osteolysis and loosening); 0.652-0.783 sensitivity and 0.976-0.988 specificity for infection; and 0.643-0.667 sensitivity and 0.894-0.939 specificity for stiffness, instability, and nonspecific pain. Diagnostic accuracy was higher in the index TKA cohort than in the revision TKA cohort. Interobserver and intraobserver reliabilities were almost perfect (κ = 0.82 and κ = 0.83, respectively). Conclusion MR imaging can help distinguish qualitative differences in the appearance of the synovium in TKA between particle-induced synovitis, infection, and nonspecific synovitis, with almost perfect interobserver and intraobserver reliability. Diagnostic accuracy is higher for index TKA than for revision TKA. © RSNA, 2016.