COMPUTED TOMOGRAPHY OSTEOARTHRITIS KNEE SCORE (COAKS) CONSTRUCTION, REPRODUCIBILITY AND POTENTIAL FOR STRUCTURAL PHENOTYPING

Computed tomography (CT) has been shown to be more sensitive identifying certain structural features of OA than radiography and MRI. The ability to acquire weight bearing CT (WBCT) at the knee is also an important current advantage over MRI that has created an opportunity to assess the entire weight bearing joint in 3-D. This could have implications for longitudinal monitoring of disease progression and the selection of participants in clinical trials according to structural phenotype. 1) To introduce the semi-quantitative CT OA Knee Score (COAKS); 2) to report initial interobserver reproducibility; and 3) to demonstrate its potential for structural disease phenotyping. 106 WBCT examinations were sourced from existing research projects at the Department of Rehabilitation Medicine, University of Kansas Medical Center, USA. A 4-point scoring system (0-3) was created for each cardinal radiographic feature of osteoarthritis (J = joint space width; O = osteophytes; C = subchondral cysts; S = subchondral sclerosis) to be assessed at each joint compartment (MTF = medial tibiofemoral; LTF = lateral tibiofemoral; PF = patellofemoral; PTF = proximal tibiofibular). Brief verbal definitions are given for each feature score in table 1. After initial training and grading of all knees by a musculoskeletal radiology clinical fellow (J.J.H.), a reference atlas was created that included a standardised review guide (including display settings and multiplanar reformat orientations), examples of each score for each feature at each compartment, and common pitfalls. Blinded to the imaging data, 10 test knees and 35 study knees were super-selected by their initial scoring to stratify disease features as evenly as possible. No restrictions were placed on the quality of imaging, the presence of ACL reconstruction (an inclusion criteria for one source study), or imaging review platform. T.D.T. trained another consultant musculoskeletal radiologist (Z.A.), who both assessed the 10 test knees, reviewed performance, then repeated the test assessment. Both then scored the 35 study knees without communication during the assessment period. A cloud score sheet was used in Google Sheets (table 2) with the reference atlas in Google Docs, both accessible online or via the respective smart phone / tablet app. The weighted kappa score for interobserver reproducibility was calculated for each structural feature across all compartments. Compartment feature scores were also plotted as heat maps for each knee to provide a visualisation of structural OA phenotype. The weighted kappa scores (+95% CI) for each feature were: J = 0.87 (0.86-0.89); O = 0.79 (0.76-0.82); C = 0.66 (0.63-0.70); S = 0.84 (0.82-0.86). On average, it took less than 5 minutes to score a single knee. Selected heat maps (figure 1) show how different phenotypes can be ascertained using the COAKS system: RUNA053L is an early MTF phenotype with PF involvement; RUNA045R is a late MTF phenotype with multicompartment disease; RUNA023L is a late LTF phenotype with severe multicompartmental osteophytosis but no other features; RUNA088L is a PF phenotype with small tibiofemoral osteophytes only. A whole joint COAKS score is also possible (out of 48), as is a score by compartment (out of 12) or by feature (out of 12) as desired. COAKS is a feasible semi-quantitative WBCT-based scoring system of knee OA that demonstrates initial substantial to almost perfect inter-observer agreement across radiographic features. The ability of COAKS to score and phenotype the weight bearing knee in 3-D may exceed individual capabilities of MRI and radiography in certain aspects, offering flexibility in phenotyping for patient selection as well as potential sensitivity for longitudinal monitoring of disease severity in clinical trials or cohort studies. It is also possible to use the COAKS in non-weight bearing CT if flagged as such and not used to compare with weight bearing studies.