Bone-RADS and OT-RADS both worked in children, with only moderate reader agreement

Design

Retrospective single-centre validation study, five readers

Population

80 patients under 18 with a solitary bone lesion on MRI

Primary outcome

Inter-reader agreement and diagnostic accuracy of Bone-RADS and OT-RADS

Effect

Fleiss κ 0.470 (0.410–0.530) Bone-RADS; 0.432 (0.389–0.475) OT-RADS

Bone-RADS and OT-RADS were developed to standardise reporting of bone lesions, largely from adult data. This Turkish study had five radiologists of differing experience apply both systems to MRI of 80 children and adolescents (age 4 to 17) with a solitary bone lesion.

Agreement across all readers was moderate for both: Bone-RADS κ 0.470 (95% CI 0.410 to 0.530) and OT-RADS κ 0.432 (0.389 to 0.475). Diagnostic accuracy for a binary benign-versus-concerning classification was moderate to good for Bone-RADS and moderate to excellent for OT-RADS. The authors credit Bone-RADS's algorithm for its slightly better agreement; OT-RADS depends more on the reader knowing each entity.

For general radiologists who see paediatric bone lesions only occasionally, a structured system gives a management category the orthopaedic surgeon can act on. Bone-RADS is the easier to apply consistently. Moderate agreement still means a meaningful share of lesions will be categorised differently by another reader, so borderline categories deserve discussion at a tumour board.

• Use a structured bone tumour system on paediatric MRI; Bone-RADS gave slightly better agreement.
• Correlate with the radiograph — matrix, margin and periosteal reaction still carry most of the diagnosis.
• Discuss intermediate categories with a musculoskeletal tumour team rather than deciding alone.
• Include a clear management recommendation (leave alone, follow up, biopsy) in the report.

Why it matters

It supports using adult-derived bone reporting systems in children, where many lesions are first reported by non-specialists.

Don't overread it

Eighty children at one centre; the accuracy ranges are broad and need external validation.

The statistics, in plain English

A Fleiss κ of 0.43 to 0.47 means readers agreed moderately beyond chance — better than guessing, but one reader in several will often choose a different category for the same lesion.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.