Fifteen independent observers of three levels of experience (consultant staff, fellows, residents) assessed 40 radiographs of children presenting with Perthes' disease using the Catterall and the Salter-Thompson grading systems. Each observer was supplied with descriptions and illustrations of the classifications and each hip was grouped by both systems by each observer. The results were statistically analysed using 'kappa' statistics. The level of interobserver agreement was higher for the Salter-Thompson system and correlated with the level of experience of the observer. Both systems can give acceptable levels of interobserver agreement, but the Salter-Thompson grouping is simpler and easier to apply in the earlier stages of the disease when treatment must be decided, and has a higher degree of reproducibility amongst more experienced observers.

Lumbosacral nerve root anomalies have been documented in the literature for over 30 years; however, no significant quantitative studies have been undertaken so far. We describe parallel studies of 100 cadaveric specimens and an equal number of metrizamide myelograms. The anatomical specimens were prepared by wide deroofing of the lumbar canal to permit precise examination of individual nerve roots and their intradural and extradural connections. The incidence of nerve root anomalies was 14%, the L5-S1 level being most commonly involved. In sharp contrast the incidence of nerve root anomalies determined by myelography was only 4%. The anomalies were classified into four groups and the diagnostic and practical implications of our findings are discussed.