Diagnostic accuracy of digital radiography for the diagnosis of osteonecrosis of the femoral head, revisited

Abstract

Background The radiographic diagnosis of osteonecrosis of the femoral head (ONFH) is challenging for radiologists. Purpose To measure the sensitivity and specificity of digital radiography for diagnosing ONFH and to evaluate the diagnostic value of the frog-leg view. Material and Methods Patients diagnosed with ONFH by magnetic resonance imaging (MRI) (n = 132) and normal controls (n = 69) were included. Two radiologists independently graded the likelihood of ONFH and subchondral fracture on radiographs. First, the anteroposterior (AP) and frog-leg radiographs were assessed as a whole. Subsequently, randomly ordered unilaterally cropped AP radiographs were analyzed. Sensitivity and specificity were measured using MRI as reference standard. A subgroup analysis for pre-collapse ONFH was performed. The effect of within-subject correlation was analyzed using a generalized-estimating-equation approach. Inter-reader agreement was assessed with κ statistics. McNemar’s test was used to determine the additional diagnostic value of the frog-leg view. Results Among the 402 hips from 201 individuals, 191 were classified as normal, 73 as pre-collapse ONFH, and 138 as post-collapse ONFH. The sensitivity and specificity ranges for ONFH were 76–89% and 77–97%, respectively. For pre-collapse ONFH, the sensitivity and specificity were 52–71% and 78–97%, respectively. The κ ranges for the likelihood of ONFH and pre-collapse ONFH were 0.59–0.66 and 0.35–0.59, respectively. The frog-leg view did not provide additional value for diagnosing ONFH. Conclusion Diagnostic performance of digital radiography for ONFH revealed low sensitivity and poor-to-moderate inter-reader agreement for diagnosing pre-collapse ONFH. There was no additional value of the frog-leg view.