Figure 1: Examples of the UCOAG for knee OA. Intra- and Inter-rater Reliability Test-retest Reliability Number of participants 9997 Males : females 34 : 6541 : 56 Age (years) 63.9 (8.0)63.6 (8.7) Body Mass Index (kg/m 2 ) 30.2 (5.5)31.3 (4.7) Table 1: Description of participant samples [mean (standard deviation)]. Table 2: Measurements of intra-rater, inter-rater and test-retest reliability for the UCOAG total score. Reliability of a unicompartmental scale for the radiographic evaluation of knee osteoarthritis: Data from the Multicenter Osteoarthritis Study (MOST) Reliability of a unicompartmental scale for the radiographic evaluation of knee osteoarthritis: Data from the Multicenter Osteoarthritis Study (MOST) L Sheehy 1, TDV Cooke 1, J Lynch 2, M Nevitt 2, L McLean 1, J Niu 3, NA Segal 4, J Singh 5, E Culham 1 1 School of Rehabilitation Therapy, Queen’s University, Kingston, ON, Canada 2 Dept. of Epidemiology and Biostatistics, University of California, San Francisco, CA 3 Boston University School of Medicine, Boston, MA 4 Dept. of Orthopaedics and Rehabilitation, University of Iowa, Iowa City, IA 5 University of Alabama at Birmingham School of Medicine, Birmingham, AB Purpose To assess the intra-rater reliability (one reader assesses a radiograph more than once), inter-rater reliability (more than one reader assesses the same radiograph), and test-retest reliability (one reader assesses two radiographs of the same person) of the UCOAG. Introduction Radiographs of the knee are commonly used to determine the severity of knee osteoarthritis (OA) and monitor its progression. Radiographs are often assessed using ordinal (semi-quantitative) scales. Assessment methods must be reliable, valid and sensitive to change. Commonly-used scales lack sensitivity to change or predominantly rely on one feature of knee OA. The unicompartmental osteoarthritis grade (UCOAG) is a composite scale that grades several individual features of knee OA in the most-affected tibiofemoral compartment and sums them to create a total score. In this scale one feature is not emphasized and all presentations of knee OA can be assessed equally. Results Participant samples are described in Table 1. As an example, the mean total score for reader 3, first reading was 4.1 (standard deviation 2.1) with a range of The median was 4.0 and the interquartile range was 2.0. ICC (2,1) and Cohen’s weighted kappa results are presented in Table 2. The MDC 90 for test-retest reliability was Methods Radiograph Selection: a)Intra-rater and Inter-rater Reliability 100 fixed-flexion radiographs were selected from the Multicenter Osteoarthritis Study (MOST) database. Radiographs were stratified according to knee OA severity as seen on magnetic resonance images (MRI). Joint space narrowing and frontal alignment were used to determine the most-affected tibiofemoral compartment (70% medial). b)Test-retest Reliability 100 fixed-flexion radiograph pairs of knees that did not change in OA severity over 15 or 30 months were selected from the MOST database. The absence of change was defined as no change in tibiofemoral cartilage morphology seen on MRI. MRI findings were used to stratify potential radiographs according to severity of knee OA. The medial tibiofemoral compartment was most- affected in 70% of the images. UCOAG Reading Procedure: 3 readers were trained and printed instructions were provided. For intra-rater and inter-rater reliability 3 readers graded each of the 100 radiographs twice, at least two weeks apart. For test-retest reliability one reader graded all 200 radiographs. Order of radiographic presentation was randomized. Analyses: Intraclass correlation coefficients (ICC 2,1 ) and Cohen’s weighted kappa were performed for each type of reliability. Minimal detectable change (MDC 90 ) was calculated using test-retest reliability data. Discussion and Conclusions The UCOAG has moderate to excellent intra-rater, inter-rater and test- retest reliability over a large range of knee OA severity. These results are comparable or better than the reliability results of other scales used to grade knee OA on a radiograph. A change of two or more grades in the UCOAG scale indicates a true change in TF OA severity. None of the 4 OA features was more or less reliable than the others and OA severity also did not affect reliability. Suggestions to improve reliability include increased training and the use of an atlas of radiographs showing the 4 knee OA features. The UCOAG is recommended to grade knee radiographs for knee OA severity, for clinical and research purposes. ReaderN ICC (2,1) (95% CI) Cohen’s Weighted Kappa (95% CI) Intra-rater Reliability Reader (0.77; 0.89) 0.65 (0.56; 0.74) Reader (0.88; 0.94) 0.75 (0.68; 0.82) Reader (0.86; 0.93) 0.70 (0.63; 0.77) Inter-rater Reliability Readers 1 vs (0.72, 0.85) 0.47 (0.37; 0.57) Readers 1 vs (0.48; 0.65) Readers 2 vs (0.54; 0.69) Test-retest Reliability Time 1 vs. 2, reader (0.80; 0.90) 0.64 (0.57; 0.72) Reference Cooke et al., Radiographic grading for knee osteoarthritis. A revised scheme that relates to alignment and deformity. J Rheumatol 1999;26(3): Acknowledgements Software for grading radiographs was provided by Orthopedic Alignment and Imaging Systems Inc. (OAISYS). MOST is funded by NIH and NIA. MOST Online ( Felson (Boston University) AG18820Lewis (University of Alabama at Birmingham) AG18947 Torner (University of Iowa) AG18832Nevitt (University of California, San Francisco) AG19069 UCOAG The worst-affected compartment is assessed. OA FeatureMaximum Grade Joint Space narrowingJS3 Femoral OsteophytesFO3 Tibial ErosionTE4 SubluxationSU3 Total ScoreTS13 Medial Compartment: JS1 FO1 TE0 SU0 TS2 Medial Compartment: JS3 FO1 TE1 SU1 TS6