COMPARISON BETWEEN THE GRAF METHOD AND A NEW 50/50 METHOD OF INTERPRETING ULTRASONOGRAMS IN DEVELOPMENTAL DYSPLASIA OF THE HIP

Abstract

We have encountered radiological reports of ‘normal Graf α-angles’ when the femoral head was subluxed. We therefore developed a simple method to determine femoral/acetabular congruency known as the 50/50 method. We compare our method to the established Graf method.

Two identical, randomly assorted sets of 100 ultra-sonograms were evaluated. All ultrasonograms were of patients under 3 months of age within our DDH screening program. The images were assessed to be either ‘normal’ or ‘abnormal’ by 6 FY1’s using each method after reading brief instructions. (Images were classified as normal or abnormal by consensus between an orthopaedic consultant and radiologist who also examined and preformed dynamic screening on each infant).

The mean proportion of abnormal scans with agreement and normal scans with agreement was 0.52 (95% CI 0.39–0.69) and 0.92 (CI 0.87–0.96) respectively, indicating moderate agreement (kappa 0.41, CI 0.12–0.71) for inter-observer variability using the Graf method. On average the inter-observer variability using the 50/50 method for abnormal and normal scans with agreement was 0.60 (CI 0.35–0.84) and 0.92 (95% CI 0.85–0.99) respectively with moderate agreement (kappa 0.50, CI 0.20–0.80). Intra-observer variability between the Graf and 50/50 methods revealed moderate agreement (mean kappa 0.41, CI 0.17–0.66) with the average proportion of abnormal and normal scans with agreement of 0.50 (CI 0.32–0.69) and 0.91 (CI 0.83–0.98) correspondingly. The accuracy of each test was equal, ranging from 84% to 93%.

The 50/50 method is straightforward to both use and teach. Moreover, it successfully serves as “red dot” system to flag up abnormal hips at clinic. The 50/50 method is at least as good as Graf with regard to accuracy, inter-observer and intra-observer variability. We recognise that dynamic screening remains the gold standard.