Which Classification System Is Most Useful for Classifying Osteonecrosis of the Femoral Head?

Background Many studies have confirmed that the size and location of necrotic lesions are major factors that affect the prevalence of collapse and prognosis in patients with osteonecrosis of the femoral head (ONFH). Although several classification systems categorize and quantify ONFH, there is no agreement on which one is most useful for the purpose. Questions/purposesWe compared the Steinberg, modified Kerboul, and Japanese Investigation Committee (JIC) classifications of ONFH in terms of (1) the correlation among the three different classification systems. We further examined (2) the inter- and intraobserver reliability of the three classification systems and (3) the association of higher grades within each classification and the risk of subsequent collapse. Methods Between January 2000 and December 2014, we treated 101 hips in 74 patients for precollapse ONFH, diagnosed either on plain radiographs or MRI. Of those, one patient (1%) died, six patients (8%) were lost to followup, and two patients (3%) underwent osteotomy before 2 years, leaving 86 hips in 65 patients (88%) for analysis here. Three-dimensional spoiled gradient-echo sequence (3D-SPGR) MRI was performed for all hips, and the presence of ONFH was determined by finding the area surrounded by the outer margin of the low-signal-intensity band on 3D-SPGR MRI. Patients with ONFH were categorized using the Steinberg, modified Kerboul, and JIC classification systems, and correlations among these three classification systems were investigated. Inter- and intraobserver reliability was assessed by 10 orthopaedic surgeons using 40 sets of 3D-SPGR MR images. The reliability of each system was evaluated using the kappa coefficient. The cumulative survival rate with collapse and undergoing hip arthroplasty as the endpoints was evaluated for each of the three classification systems (mean followup, 9 years; range, 2-16 years), and the association of higher grades within each classification and the risk of subsequent collapse were also evaluated. Results We found strong correlations between the Steinberg and modified Kerboul classifications (rho= 0.83, p < 0.001), the Steinberg and JIC classifications (rho= 0.77, p < 0.001), and the modified Kerboul and JIC classifications (rho= 0.80, p < 0.001). Interobserver reliability in the JIC classification (0.72; range, 0.30-0.90) was higher than that in the Steinberg classification (0.56; range, 0.24-0.84; p < 0.001) and the modified Kerboul classification (0.57; range, 0.35-0.80; p < 0.001). The cumulative survival rate with collapse as the endpoint after a minimum of 2 years of followup in the Steinberg classification differed between Grades A (82%; 95% confidence interval [CI], 66%-97%) and B (43%; 95% CI, 21.9%-64.8%; p = 0.007), Grades A and C (20%; 95% CI, 4.3%-35.7%; p < 0.001), and Grades B and C (p = 0.029). Survival was lower for modified Kerboul Grade 4 hips (12%; 95% CI, 0%-27.1%) than for Steinberg Grade C hips (20%; 95% CI, 4.3%-35.7%) and JIC Type C2 hips (18%; 95% CI, 2.8%-34.0%). The JIC classification was best able to identify hips at low risk of collapse because no JIC Type A hips collapsed. Conclusions The JIC classification was more reliable and effective, at least for early-stage ONFH, than the Steinberg or modified Kerboul classifications. Further investigation might be useful to identify whether each classification system emphasizes specific risk factors for collapse.Level of EvidenceLevel III, diagnostic study.