Reliability of detection of ultrasound and MRI features of hand osteoarthritis: a systematic review and meta-analysis

OBJECTIVES: To systematically review the literature on inter- and intra-rater reliability of scoring US and MRI changes in hand OA. METHODS: MEDLINE, EMBASE, CINHAL, Web of Science and AMED were searched from inception to January 2020. Kappa (κ), weighted kappa (κ(w)) and intra-class correlation coefficients for dichotomous, semi-quantitative and summated scores, respectively, and their 95% CI were pooled using a random-effects model. Heterogeneity between studies was assessed and reliability estimates were interpreted using the Landis–Koch classification. RESULTS: Fifty studies met the inclusion criteria (29 US, 17 MRI, 4 involving both modalities). The pooled κ (95% CI) for inter-rater reliability was substantial for US-detected osteophytes [0.66 (0.54, 0.79)], grey-scale synovitis [0.64 (0.32, 0.97)] and power Doppler [0.76, (0.47, 1.05)], whereas intra-rater reliability was almost perfect for osteophytes [0.82 (0.80, 0.84)], central bone erosions (CBEs) [0.83 (0.78, 0.89)] and effusion [0.83 (0.74, 0.91)], and substantial for grey-scale synovitis [0.64 (0.49, 0.79)] and power Doppler [0.70 (0.59, 0.80)]. Inter-rater reliability for dichotomous assessment was substantial for MRI-detected CBEs [0.75 (0.67, 0.83)] and synovitis [0.69 (0.51, 0.87)], slight for osteophytes [0.14 (0.04, 0.25)], and almost perfect for sum score of osteophytes, CBEs, joint space narrowing (JSN), and bone marrow lesions (BMLs) (0.81–0.89). Intra-rater reliability was almost perfect for sum score of MRI synovitis [0.92 (0.87, 0.96)], BMLs [0.88 (0.78, 0.98)], osteophytes [0.86 (0.74, 0.98)], CBEs [0.83 (0.66, 1.00)] and JSN [0.91 (0.87, 0.91)]. CONCLUSION: US and MRI are reliable in detecting hand OA features. US may be preferred due to low cost and increasing availability.