Mylène Jansen

KIDA performance in severe OA 225 11 Table 5 : Influence of mean values and Kellgren-Lawrence grade on the intra-observer differences between measurements for severe radiographs analyzed within 1 month Mean value Kellgren-Lawrence grade B β P- value B β P- value Bone density (all in mm Al eq) Femur mean lateral 0.002 0.078 0.443 0.024 0.176 0.083 Femur mean medial 0.006 0.186 0.066 -0.013 -0.081 0.427 Tibia mean lateral 0.032 0.159 0.118 -0.060 -0.057 0.575 Tibia mean medial 0.011 0.223 0.021 0.025 0.100 0.326 JSW (all in mm) Mean -0.028 -0.098 0.336 -0.025 -0.081 0.428 Mean lateral 0.011 0.032 0.755 -0.066 -0.104 0.309 Mean medial -0.006 -0.067 0.515 0.016 0.093 0.365 Minimum 0.070 0.314 0.002 0.008 0.029 0.778 Osteophytes (all in mm 2 ) Femur lateral 0.050 0.173 0.089 0.156 0.072 0.479 Femur medial 0.130 0.244 0.016 1.340 0.271 0.007 Tibia lateral 0.257 0.577 <0.001 1.783 0.325 0.001 Tibia medial 0.393 0.606 <0.001 1.375 0.285 0.004 Other (mm, mm, degrees) Eminence lateral 0.043 0.114 0.264 0.134 0.141 0.167 Eminence medial 0.021 0.059 0.564 -0.040 -0.059 0.565 Joint angle -0.011 -0.048 0.639 -0.069 -0.084 0.413 Separate linear regression models were used the mean value and the Kellgren-Lawrence grade, for all different parameters. β : standardized coefficient; B : unstandardized coefficient; mm Al eq: mm aluminum equivalent. Discussion Based on the presented results, it is shown that KIDA is a useful tool for radiographic analysis of OA characteristics even in patients with severe OA. Notably, (re)analyzing images in a short time period increases reproducibility (decreases SDDs and CVs). Furthermore, the systematic bias between measurements decreases when images are reanalyzed within a short time period (1 month compared to years). This emphasizes the importance of performing the analyses required for a specific research question within a limited time period and randomized for time/ visit sequence. The fact that some parameters showed significant differences between readings, even for the images reanalyzed within 1 month, can only be speculated on. For these parameters, the direction of this bias was the same for the images that were analyzed over years and over months, which implies that the bias is expectedly systematic and not coincidental (i.e. not because of subtly different conditions that may unconsciously affect measurement) and that it is not likely the result of recalling the first reading. For most consistent biases, the direction was positive. As such, changes over time for bone density, femoral osteophytes, and lateral JSW might be overstated in case images are analyzed in chronological order of acquisition over time