The final cohort consisted of 285 patients (150 males, 135 females) with a mean age of 49.0 years (SD 21.0; range 8-92).The most frequent histological diagnosis of the bone lesions was metastasis (72/285, 25.3%) followed by atypical cartilaginous tumor (ACT)/chondrosarcoma grade I (54/285, 18.9%) - (Table 3).

Interobserver agreement (Table 4 - Figure 1)

The two musculoskeletal Radiologists (Readers 1 and 2) showed at least substantial agreement for all the Bone-RADS categories (Kappa values ranging from 0.84 to 0.99).

The highest agreement was observed in the assessment of pathological fractures (Cohen’s Kappa: Readers 1-2 = 0.99, Readers 1-3 = 0.92, Readers 2-3 = 0.93).

The lowest Kappa values were found for the “endosteal scalloping” category when comparing the general radiologist with the two musculoskeletal radiologists (K=0.59 for Readers 1-3 pairing;  K=0.55 for Readers 2-3 pairing).

Agreement on the final Bone-RADS score was substantial/almost perfect for all the three pairings (Kappa values ranging from 0.79 to 0.84).

Diagnostic performance (Table 5 - Figure 2)

The score showed good diagnostic performance in discriminating between benign and potentially aggressive lesions (AUC=0.82).

Suboptimal sensitivity (74.9%) and NPV (50.5%) were found using a Bone-RADS≥3 cut-off (intermediate-risk). This may partly be explained by the high prevalence in our cohort of atypical cartilaginous tumors (ACT), which are intermediate lesions, often difficult to differentiate from benign lesions such as enchondromas. In our cohort, 46 out of 53 cases of ACT yielded a Bone-RADS score = 2.

If enchondroma and atypical cartilaginous tumors were excluded, diagnostic performance significantly improved (AUC=0.92; Sn=95.0%; NPV=86.4%).

In figure 3, 4 and 5, examples of bone lesions evaluated through the ACR Bone-RADS are provided.