Collapse-Related Bone Changes in Osteonecrotic Femoral Heads at Multidetector CT: Comparison between Femoral Heads with Limited and Advanced Collapse

1- Patients’ population

The Institutional Review Board approved the study, with the informed consent of the patients being waived because the study was performed on femoral head specimens that were resected for total hip replacement. An MSK radiologist with four years of experience collected 14 femoral head (FH) specimens with osteonecrosis (ON) from 13 patients: 8 men (46.9 years ±13.9) and 5 women (68.4 years ±16.05). These specimens were resected during hip replacement. The FH specimens were fixated in formalin. In a preliminary study on eight FH specimens, the sensitivity and specificity of in vitro MDCT were calculated in comparison to µCT (see online supplemental material).

2- Staging of ONFH and measurement of articular surface collapse

Staging of ONFH was performed on radiographs and MRI according the ARCO classification [4]. Collapse of the articular surface was measured using the best fitting concentric circle technique [5]. The median collapse was 1.5 mm [1.1–2.1 mm, 95% CI]. There were 10 stage 3-As, 3 stage 3-Bs, and 1 stage 4 ONFH.

3- Data acquisition and grid overlying

Fourteen resected FH specimens were scanned on a 40-row MDCT (Siemens Somatom Definition 40, Erlangen, Germany). One radiologist with four years of experience viewed the MDCT images on an image viewer with multiplanar capacity. Coronal reformats were obtained and segmented by overlaying a transparent 12 × 12 grid. The anonymized images with the overlaid grid were uploaded on an in-house developed software that enables image analysis (Figure 1 and E1). Further details are provided in online supplemental material.

Figure 1

4- Image analysis and lesion definition

Images were independently analyzed by an MSK radiologist with 30 years of experience (R1), and an MSK radiologist with four years of experience (R2). The reading process was performed in 4 × 4 mm grid boxes using the grid overlay method on in-house developed software (see online supplemental material).

Bone changes at MDCT were defined as follows: interface-related trabecular sclerosis corresponded to thickened trabeculae with a band-like distribution and a concave orientation towards the articular surface. Cortical bone interruption corresponded to a focal interruption of the subchondral bone plate with or without deformity (Figure 2). Trabecular bone interruption corresponded to an interruption of two or more contiguous trabeculae, having irregular, sharp, or angular margins with or without a gas-filled cleft or a dense linear band corresponding to trabecular crushing (Figures 2, 3). Trabecular bone resorption corresponded to a lucent zone devoid of mineralized content, with smooth or rounded margins (Figure 2).

Figure 2

Figure 3

Limited FH collapse was defined by the presence of a depression of the FH contours <1.5 mm, and advanced FH collapse by the presence of a depression ≥1.5 mm.

5- Statistical analysis

Data output was automatically generated on Excel sheets. Frequency of grid boxes with bone changes was compared between FH with limited and advanced collapse using Chi-squared test. Interobserver agreement was computed using Cohen’s Kappa and interpreted according to the scale proposed by Altman [6]. A p-value <0.05 was regarded as statistically significant. Statistical analyses were performed by using MedCalc Statistical Software version 19.2.6 (MedCalc Software bv, Ostend, Belgium; https://www.medcalc.org; 2020).

1- Frequency of grid boxes with bone changes at MDCT

There were 2,933 grid boxes containing cortical bone and 10596 grid boxes containing trabecular bone on the MDCT reformats. For R1, there were 557/2933 (19%) with cortical bone interruption, 796/10596 (7.5%) with trabecular bone interruption, and 331/10596 (3.1%) with trabecular bone resorption. Results for R2 are given in Table 1.

2- Topology of grid boxes with necrotic lesion and collapse-related bone changes at MDCT

For R1, grid boxes with collapse-related bone changes (CRBC) at MDCT were more frequent in the anterior, central, and superior thirds of the FH (Table 2). Grid boxes with trabecular interruption were more frequent in the superficial (448/2621; 17.1%) than in the deep layer (348/7975; 4.4%). Grid boxes with trabecular resorption were more frequent in the deep (281/7975; 3.5%) than in the superficial layer (50/2621; 1.9%). Detailed results for R1 and R2 are provided in Table 2.

For R1, grid boxes with cortical and trabecular bone changes were more frequent within than outside the necrotic lesion (p < 0.001). Detailed results for R1 and R2 are provided in Table 3.

3- Comparison of frequency of grid boxes with necrotic lesions and CRBC at MDCT between limited and advance collapse

For R1, grid boxes with cortical interruption were more frequent in FH with collapse ≥1.5 mm (20.49%) than in FH with collapse <1.5 mm (17.02%) (p = 0.018). There was no statistical difference in the frequency of grid boxes with trabecular interruption between FH with collapse ≥1.5 mm (7.13%) than <1.5 mm (8%) (p = 0.094). Grid boxes with trabecular resorption were more frequent in FH with collapse ≥1.5 mm (3.88%) than <1.5 mm (2.15%) (p < 0.001). Detailed results for R1 and R2 are provided in Table 4.