Are the Sanders-Frykberg and Brodsky-Trepman Classifications Reliable in Diabetic Charcot Neuroarthropathy?

The purpose of this study was to assess the intra- and inter-reader reliability of the 2 Charcot neuroarthropathy classifications (Sanders-Frykberg and Brodsky-Trepman), as well as Eichenholtz staging. We hypothesized that the inter-reader reliability, with respect to these 3 classification systems, would be moderate at best. Digital radiographic images were organized in a digital slide presentation without clinical information. All 5 reviewers underwent a standard training session administered by the principal investigator, reviewing 5 cases of Charcot neuroarthropathy. Images of 55 cases of Charcot neuroarthropathy and 5 normal cases were distributed to each of the 5 physicians electronically, who independently rated all 60 cases according to the 3 classification systems. The 95% confidence interval of the intraclass correlation coefficient estimate for Sanders-Frykberg was 0.9601 to 0.9833 at week 0 and 0.9579 to 0.9814 at week 8, which can be regarded as "excellent" reliability. For Trepman-Brodsky, the 95% confidence interval of the intraclass correlation coefficient estimate was 0.8463 to 0.9327 at week 0 and 0.8129 to 0.9226 at week 8, which can be regarded as "good" to "excellent" reliability. For Eichenholtz, the 95% confidence interval of the intraclass correlation coefficient estimate was 0.6841 to 0.8640 and 0.6931 to 0.8730 at weeks 0 and 8, respectively, which can be regarded as "moderate" to "good" reliability. The classification systems of Charcot neuroarthropathy are an important tool for communication among physicians. Based on the results at our institution, the Sanders-Frykberg classification exhibited the best inter-reader performance. The Trepman-Brodsky classification exhibited good to excellent reliability as well. The intraclass correlation coefficient of the Eichenholtz classification was moderate to good.