In a prospective study, 232 neonates were examined sonographically using the methods of Graf and Terjesen. In order to determine the reproducibility of the methods, 50 hips were evaluated by two skilled examiners. In an inter-observer study, five physicians and five medical students evaluated 24 images, which were evaluated on ten occasions at twoweekly intervals by one of the authors. Statistical evaluation used the Bland-Altman approach. The neonates (110 females, 122 males) were less than

more »

... r days old. The mean α α α α angle was 62.4˚ and mean femoral head cover was 55.4%. According to Graf's method, 1.3% of hips were pathological, compared with 4.1% according to Terjesen. Spearman's correlation coefficient between femoral head cover and α α α α angles was 0.552. The Bland-Altman approach shows greater variation for femoral head cover than for α α α α , if measured by experienced examiners. The Bland-Altman approach shows almost equal reproducibility for α α α α and femoral head cover in the inter-observer test, but better repeatability for α α α α in the intra-observer test. The Graf results relate better than Terjesen's to the well-known frequency of 1% to 2% hip dysplasia in the European population. Kappa statistics indicate a fair agreement between the two methods. Inter-observer evaluation shows an equal reproducibility of both methods, whereas intra-observer tests reveal better repeatability with Graf's method. The foundations for the sonographic diagnosis of hip dysplasia in the newborn were described by Graf, 1 and his method was rapidly established in the German-speaking countries. In the English-speaking world and Scandinavia, Harcke et al 2 and Terjesen, Rundén and Tangerud 3 and Terjesen, Bredland and Berg 4 developed their own method of investigation. Whereas Graf's method measures two angles, α and β, Harcke et al 2 and Terjesen et al 3,4 determine the amount of the femoral head covered by the osseous part of the acetabulum. Ultrasound measurement of anatomical specimens in a water bath gave comparable reproducibility for the two methods, 5,6 but so far there have been few such clinical studies. This study aimed to compare the Graf and Terjesen methods and to clarify whether, under clinical conditions, differences in the reproducibility of the methods can be assessed. Intra-and inter-observer variability is demonstrated by different examiners with variable experience in hip sonography and by repeated evaluations by one of the authors (DS). Materials and Methods In a prospective study between February and September 2003, 232 consecutive neonates were investigated sonographically by the methods of Graf and Terjesen. The examinations were performed with a 7.5 MHz linear transducer. The images were taken such that the landmarks of both methods could be identified. Documentation was by a video printer (Sony Video Graphic Printer UP 890 MD; Sony, Tokyo, Japan) with a magnification factor of 1.4 (Fig. 1 ) and the images were evaluated as follows. Correlation between Graf and Terjesen. Direct comparisons of the α angle and the linear variables a and b for the determination of femoral head cover were made by one examiner from the same image. We did not evaluate the β angle because it is not relevant in normal hips (more than 98% of our hips) and we wished to limit our data. Correlation analysis used Spearman's coefficient because the data were not parametric. In order to assess agreement between the two methods with respect to clinical outcome, the Graf type was considered