Validation of the Bone-RADS (Bone Reporting and Data System) Score for Predicting Risk of Malignancy

Bone-RADS (Bone Reporting and Data System) is a consensus guideline by the American College of Radiology to assist in the risk stratification of lytic bone lesions. A score is calculated using 5 radiographic factors and a history of primary malignancy. The purpose of this study was to externally validate the Bone-RADS scoring system.

This retrospective review examined radiographs from January 1, 2021, to December 31, 2021. Radiographs were identified via a keyword search for "lucent," "osteolytic," or "lytic" in extremity radiograph reports. All radiographs were initially read by a musculoskeletal radiologist. Radiographs of skeletally immature patients, lesions in a bone other than a long bone, primary soft-tissue masses, and negative/normal radiographs were excluded. Radiographs were independently scored on the basis of the Bone-RADS system by 4 reviewers. In the absence of a biopsy, a lesion was considered benign if it remained stable for 2 years on radiographs. Scores were correlated with final diagnoses to calculate the rates of malignancy and malignant transformation.

Of 1,108 screened radiographs, 423 were eligible. The mean age of the patients was 62 years and 49% were male. A higher Bone-RADS score was associated with a malignant diagnosis (p < 0.001), with an odds ratio of 9.7 (95% confidence interval [CI], 6.2 to 14.9) per 1-unit increase in the score. A higher margin score (p < 0.001), endosteal erosion (p < 0.001), a pathologic fracture (p < 0.001), and a soft-tissue mass (p = 0.02) were associated with increased risk of malignancy. The mean inter-rater agreement (weighted kappa) was 0.52, the inter-rater mean absolute deviation (MAD) was 0.37, and the intra-rater MAD was 0.27. Sensitivity was 97.1% (CI, 94.4% to 98.5%), specificity was 67.4% (CI, 59.3% to 74.5%), positive predictive value was 85.2% (CI, 80.9% to 88.7%), and negative predictive value was 92.4% (CI, 85.7% to 96.1%).

The Bone-RADS scoring system demonstrated moderate diagnostic performance for malignancy risk stratification of lytic bone lesions, with high sensitivity and reproducible binary classification of possibly benign versus malignant. Although inter-rater agreement for the 4-tier score was only moderate, clinically relevant benign-versus-malignant categorization demonstrated higher inter- and intra-rater agreement. These findings support Bone-RADS as a practical triage framework based on musculoskeletal imaging of lytic bone lesions.

Diagnostic Level III. See Instructions for Authors for a complete description of levels of evidence.