Twenty observers reported independently on the presence or absence of a fracture of the scaphoid on 60 sets of radiographs; these included initial and 2- to 3-week views in patients in whom the outcome was known, normal scaphoids and random copies of these. Analysis of variance of the accuracy of observations revealed that the 2- to 3-week radiographs did not improve diagnostic ability and that this was independent of the experience or seniority of the observer. For normal radiographs, 20% of the observations reported a fracture. Reproducibility of opinion improved with experience but this did not help with accuracy. Radiographs without accurate clinical observation should not determine the management of the suspected scaphoid fracture.

Inter-observer agreement and reproducibility of opinion were assessed for the radiographic diagnosis of union of scaphoid fractures on films taken 12 weeks after injury. Weighted kappa statistics were used to compare the opinions of eight senior observers reviewing 20 sets of good quality radiographs on two occasions separated by two months. There was poor agreement on whether trabeculae crossed the fracture line, whether there was sclerosis at or near the fracture and on whether the proximal part of the scaphoid was avascular. As a consequence, agreement on union also was poor; it appears that radiographs taken 12 weeks after a scaphoid fracture do not provide reliable and reproducible evidence of healing.