Background/Purpose:   Assessment of structural progression of knee osteoarthritis (OA) requires quantitative evaluation of the radiographic joint space width (JSW). Significant reduction in the rate of narrowing of JSW represents a crucial readout in the evaluation of potential disease modifying agents for OA. However, as the knee is geometrically complex, quantitative JSW readings can only be performed on radiographs obtained using special techniques, and even in such cases, the variability in measurements (both intra-observer and inter-observer) is often large. Thus, to confidently demonstrate significance, longitudinal OA studies often require very large group sizes to achieve satisfactory power. Reducing observer variability at baseline could be a useful strategy to permit smaller or shorter structural OA studies. It has long been clear that observer variability is observer-related. However, we observed that the geometry of individual knee joints often predisposes to exagerrated measurement variability. Here we tested the hypothesis thatthere is a study subject–related component to observer variability of JSW mesurements, and that this can be assessed at baseline to minimize the risk of observed variability in longitudinal studies.

Methods:   With IRB approval and after obtaining informed consent, 20 subjects with symptomatic medial compartment knee OA (Kellgren-Lawrence grade 2-3, pain on ambulation >30 mm on a 100 mm visual analog scale) were evaluated at baseline and longitudinally as part of a larger trial. Subjects underwent semi-flexed fluoroscopic-guided PA knee radiography (Schuss view) at baseline and at 12 months. Medial compartment JSWs were quantified using Image J software (US NIH, Bethesta, MD, http://rsb.info.nih.gov/ij/) twice by the same observer, and these measurements were repeated after 8 years. The difference in the measured medial JSWs between the measurements made at baseline and those repeated after 8 years of the baseline x-ray for each subject was calculated (baseline observer variability for the individual subject). Similarly, the difference between two measurements of 12-month xrays were calculated for each subject (follow-up observer variability). Pearson’s correlation test was used for analysis.

Results:  Baseline observer variabilities ranged from -1.55 to 0.88, with a mean ±SD of -0.05 ± 0.44. Similarly, the variability of the 12 month measurements ranged from -1,13 to 1.35, with a mean ±SD of 0.13 ± 0.44. The discrepancy at baseline was related to the measured discrepancy of the subjects at 12 months, (Pearson’s r=0.51, p=0.02), and these measurements were consistent when re-assessed 8 years later (intraclass correlation coefficient 0.97 (95 % CI 0.93-0.99)).

Conclusion:  These data suggest that some knees are subject to greater measurement variability than other knees, and that these differences remains constant over time. It may be possible to identify knees that are highly prone to measurement error at baseline, and to thereby reduce groups sizes in OA trials by excluding such knees. . Selecting patients with radiographs resulting in better observer variabilty at baseline might allow performing disease modifying OA studies with smaller number of patients or shorter duration.

« Back to 2016 ACR/ARHP Annual Meeting

ACR Meeting Abstracts - https://acrabstracts.org/abstract/observer-variability-of-joint-space-with-measurements-is-subject-related-in-medial-knee-osteoarthritis/