# Multicenter reliability study of the universal long bone nonunion classification

## TL;DR

This study shows that a new classification system for long bone nonunions is reliable and more effective than existing systems when used by multiple experts.

## Contribution

The study introduces and validates a new classification system for long bone nonunions with higher reliability than existing frameworks.

## Key findings

- The ULBNC system showed substantial to almost perfect inter-observer agreement for nonunion classification.
- Including clinical features like pathological mobility significantly improved agreement compared to the Weber–Cech system.
- Refining classification criteria improved group-level agreement from moderate to substantial.

## Abstract

A universal classification providing a clinically relevant and anatomically comprehensive framework for long bone nonunions has been recently introduced. This study aimed to evaluate its inter- and intra-observer reliability and compare its performance with the widely used Weber–Cech classification.

This multicenter, three-stage validation study included 191 cases meeting the FDA definition of nonunion. Four expert raters participated. In Stage 1, cases were classified using existing systems: AO/OTA for anatomical location and Weber–Cech for biological type. In Stage 2, 133 eligible cases were independently classified using the Universal Long Bone Nonunion Classification (ULBNC) in two rounds, two weeks apart. Stage 3 involved refinement of classification criteria based on feedback and statistical analysis, followed by re-assessment of 90 cases. Inter-observer reliability was assessed using free-marginal Fleiss' kappa; intra-observer reliability using Cohen's kappa with linear weighting.

Substantial to almost perfect inter-observer agreement was observed for type classification (κ = 0.85), with the highest reliability in diaphyseal nonunions (κ = 0.90). Incorporating pathological mobility significantly improved agreement compared to Weber–Cech (κ = 0.38, p < 0.05). Periarticular nonunions showed substantial agreement (κ = 0.72). Group-level agreement improved from moderate (κ = 0.42–0.57) to substantial (κ = 0.79–0.82) after refining criteria. Subgroup agreement was excellent (κ = 0.89–1.00). Intra-observer reliability ranged from substantial to almost perfect across all levels.

ULBNC is a reliable and reproducible classification system for long bone nonunions. Incorporation of clinical features—such as pathological mobility, alignment, and correction strategy—enhances its clinical utility and supports standardization in treatment and research.