- Design
- Single-centre retrospective validation study with five readers of differing experience
- Population
- 80 patients under 18 with MRI for a solitary bone lesion, January 2015 to October 2025; mean age 11.95 years
- Primary outcome
- Interrater reliability (Fleiss and Cohen kappa) and diagnostic accuracy of Bone-RADS and OT-RADS
- Effect
- Fleiss kappa 0.470 (95% CI 0.410-0.530) for Bone-RADS, 0.432 (0.389-0.475) for OT-RADS; accuracy moderate to good and moderate to excellent respectively
Bone-RADS and OT-RADS were built to standardise reporting of bone lesions and steer management. Neither was developed on children, whose normal marrow, physes and benign lesion spectrum differ enough that transferring an adult lexicon is not automatic.
This study took every patient under 18 who had an MRI for a solitary bone lesion at one centre between January 2015 and October 2025 — 80 children, mean age 11.95 years, 66 of them boys — and had five radiologists of differing experience apply both systems, in their original categorical form and as binary classifications.
Overall agreement across all five readers was moderate for both: Fleiss kappa 0.470 (95% CI 0.410-0.530) for Bone-RADS and 0.432 (0.389-0.475) for OT-RADS. Accuracy on binary classification ranged from moderate to good for Bone-RADS and moderate to excellent for OT-RADS. The authors attribute Bone-RADS's better agreement to its explicit algorithm; OT-RADS has no comprehensive lexicon and requires the reader to know each disease.
That trade is the useful part. A system with an algorithm produces more consistent reports from readers of mixed experience; a system relying on disease knowledge can reach higher accuracy in expert hands. Which you want depends on who is reading in your department at 8pm.
- Both systems can be applied to paediatric MRI — that is the question this answers
- Prefer Bone-RADS where readers vary in experience; its algorithm drives the consistency
- Moderate kappa means two readers will still disagree often; state the category, do not treat it as settled
- Single centre, 80 patients, 66 boys — a narrow sample for a validation
- Neither system substitutes for discussion at a bone tumour meeting
Why it matters
It answers whether reporting systems built for adult bone lesions can be applied to children at all.
Don't overread it
Single-centre validation in 80 children; this establishes feasibility, not that using the systems changes management or outcome.
The statistics, in plain English
A Fleiss kappa of 0.47 is conventionally 'moderate' — it means agreement beyond chance, but far from reliable. In practice two radiologists applying Bone-RADS to the same child's MRI will often assign different categories. That is not a reason to discard the system; an unstructured report agrees even less. It is a reason not to treat the category as a hard number. 'Moderate to excellent' accuracy is a range across readers, not a single estimate.
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.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for radiology, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free