The validation of a new classification of the external fixator screw-bone interface. Screw loosening significantly affects the stability of an external fixator, however radiographs are normally taken to assess bone healing and not screw loosening. This study was performed to assess the inter and intra-observer reliability of radiographic features of external fixator screw loosening. Eight observers were shown plain radiographs of 120 external fixator bone-screw interfaces on two occasions, and were asked to grade the screws according to the following features.
Solid screw. Periosteal reaction around the screw. Area of lucency around the screw. Marginal corticalisation around the screw. Frank loss of position of the screw. The overall kappa value for this study was 0.29, with the component values ranging from 0.15 to 0.41. To determine if the reliability could be improved, two observers classified 192 digitised radiographs of external fixator screws. On the first occasion the radiographs were shown at a size, brightness and contrast equal to the original film. On the second occasion the radiographs were subjected to image enhancement and magnification. This showed improvement in all the kappa values, the overall value increasing to 0.39, with similar improvements in the component parts. Unfortunately no observations were made of loose screws, therefore, two observers were asked to classify 160 digitised images of screws which were selected with a bias in their outcome, to remain solid or become loose. The observers obtained a kappa value >
0.50 for loose screws. A classification system for the bone-screw interface is of value both in research and clinical practice. Despite the fact that standard radiographic views were used the classification system described shows satisfactory inter and intra-observer reliability and this improved when digital enhancement was applied.
