LAUSR

OBJECTIVE To determine the extent of agreement between the original CSI and the Rasch analyzed version (RC-CSI) and to explore the association of both versions with psychophysical tests and their respective sensitivity and specificity. METHODS Patients with knee OA enrolled in a multicenter cohort study in Montreal, Canada completed the original CSI and RC-CSI and psychophysical tests i.e., Pressure pain thresholds (PPTs), Temporal Summation, Conditioned pain modulation using standardized protocols. Bland-Altman analyses assessed the agreement between the original CSI and the RC-CSI; Spearman correlations and chi-square analysis evaluated the association between the two CSI scores and psychophysical tests. A CSI cut point of 40 and the RC-CSI cut-point of 31.37 was used. Receiver Operating Characteristic curves and the resulting sensitivity and specificity with psychophysical tests was also analyzed. RESULTS 293 participants were included (58.7% female, mean age of 63.6 and BMI 31.9 kg/m2). The original CSI & RC-CSI mean difference 3.3/8.2, t (292)= 8.84(p < 0.001) was significantly different indicating a small bias. Small but significant inverse correlations were found for the original CSI and RC CSI scores with PPTs at the forearm and patella with variance explained ranging from 0.01 to 0.12. The largest area under the curve suggested cut-points of 23 (CSI) and 25 (RC-CSI) with 80.9% sensitivity and 38.5% specificity. CONCLUSIONS Due to poor variance explained with QST and high false positive rates, our results indicate there is little clinical value of using either version of the CSI in people with knee OA.