Reliability and reproducibility of classification systems for Legg-Calvé-Perthes disease: a systematic review of the literature.

Several classification systems are in use for Legg-Calvé-Perthes disease. Three of them: Catterall, Salter Thompson and Herring (Lateral Pillar) are most commonly used. There has been debate on which is most reliable. The purpose of this paper was to systematically analyse the literature when the classifications were compared. The Ovid (Medline) Database was used and the MeSH terms Perthes Classification and Reliability were inserted. Eleven studies were retrieved but only five were suitable for analysis as they attempted to compare the classifications. Most studies used kappa agreements as the principal outcome measure, although intraclass coefficients and percentage agreements were also used. Only four studies assessed for both intraobserver reproducibility and interobserver reliability. A further study from the references appendages was found to be suitable, and was included in the analysis. Kappa ranged from poor to fair (Salter Thompson), fair to moderate (Catterall) and moderate to good (Herring). The outcome from Legg-Calvé-Perthes disease is extremely variable. Inconsistent interpretation of the plain films may explain this, although it is likely this is multifactorial. The papers in this study show that on balance, the Lateral Pillar classification was most reliable, probably secondary to ease of use. A persistent theme was that the subchondral fracture line in the Salter Thompson system was difficult to interpret and not always present. It also showed that whilst reliability and reproducibility tended to improve with experience, disagreement was not always restricted to more junior personnel. Each classification has its merits but reliability and reproducibility remains unsatisfactory. Digital technology in the future may help delineate the lesions better and improve agreement.