Comparison of observer variation in conventional and three digital radiographic methods used in the evaluation of patients with adolescent idiopathic scoliosis

Study Design. This study is a reliability analysis of coronal Cobb angle measurements in adolescent idiopathic scoliosis obtained by multiple observers. Objective. We sought to quantify and compare the interobserver reliability of conventional radiographs and 3 methods of digital radiography. Summary of Background Data. The use of digital radiography for the evaluation of adolescent idiopathic scoliosis is being widely adopted. Previous studies comparing manual and computer-based measurements have found excellent intraobserver reliability for both techniques. Interobserver reliability of computer-based measurements on digital radiographs has not been compared with manual measurements on conventional radiographs. Other commonly used forms of output of digital radiography have not been studied. Methods. Preoperative standing posteroanterior full-length spine radiographs from 40 patients with adolescent idiopathic scoliosis were examined by 4 observers. Patients were divided into 2 groups of 20 patients. In 1 group, radiographs were obtained by conventional technique. In the other group, radiographs were obtained using a digital radiography system. Three types of output of the identical image obtained by a digital radiography system were examined, including computer-based image, printing of the image fitted onto a single film, or printing of the image onto 2 unstitched films. The Cobb angle, upper vertebra, and lower vertebra of the major curve were measured by each observer. Interobserver reliability for each technique was calculated by intraclass correlation coefficient and interobserver variance. Results. Interobserver reliability as described by intraclass correlation coefficient and interobserver variance was excellent ( 0.93 - 0.98) for measurements made on conventional, computer-based, and fitted printed radiographs. The intraclass correlation coefficient was good ( 0.87) in measurements obtained on radiographs printed on 2 unstitched films. Conclusion. Measurements made on conventional and digital radiographs using manual and computer-based techniques have similar good to excellent interobserver reliability. Interobserver reliability was lower for digital radiographs when printed onto 2 unstitched films. The results suggest that different observers will obtain similar measurements when viewing the same image, but care should be taken when interpreting images printed on 2 unstitched films.