Interobserver and intraobserver variability in radiographic assessment of osteolysis

This study quantified the variability associated with diagnosing periprosthetic osteolysis from the radiographs of total hip arthroplasty patients. Four joint arthroplasty surgeons independently assessed radiographs of 60 patients for evidence of osteolysis in different zones. The surgeons agreed on the presence of lesions in at most 57% of the zones. κ coefficients, used to quantify the extent of agreement among the surgeons, denoted poor interobserver reproducibility (κ =.28 to.44). Intraobserver reliability—determined by comparing 2 reviews of the same radiographs done by 1 surgeon 2 weeks apart—was moderate to excellent (κ =.48 to.84). We also compared the results from the most recent radiograph with those from a time series. Agreement improved when a series was reviewed. Reliable comparisons cannot be made with osteolysis rates reported by different observers. In the research setting, osteolysis rates are more reliable if they are determined by a single reviewer whose intraobserver variability has been reported. In assessing a patient for osteolysis, it is more accurate to analyze a series of radiographs than the most recent radiograph. Copyright 2002, Elsevier Science (USA). All rights reserved.