Background: Lateral condyle fractures account for 15% to 20% of pediatric elbow fractures. Among numerous proposed classification systems, the Song classification appears the most comprehensive. The utility of any classification system relies on its ability to be descriptive, reproducible, and to guide prognosis/treatment. We assessed the Song classification by applying it to 736 retrospectively treated patients.
Methods: A total of 736 pediatric patients with lateral condyle fractures were identified between 2007 and 2014. In total, 60 patients were selected for a radiographic interclass and intraclass correlation study. Radiographs of the patients were reviewed by 6 observers, who independently measured radiographs for displacement on radiographs and assigned a Song classification. Treatment and outcomes were then reviewed on all 736 patients and evaluated as a successful outcome when achieving a healed fracture at discharge without significant complication or necessitating a change from initial treatment modality.
Results: Weighted κ values for intrarater and interrater reliability to assign Song classification indicated excellent agreement. Intraclass correlation coefficients of 6 observers measuring displacement on radiographs in millimeters indicated good to excellent agreement. In total, 106 Song 1 fracture were primarily treated by casting alone and only 5.5% required conversion to operative intervention. Overall, 139 Song 2 fractures were treated by closed treatment (n=114, 82% successful nonoperatively, 16% converted to operative management) or surgical means (n=25, 100% success) without treatment superiority (P>0.999) and both modalities had high success rates. Song 3 fractures (n=17) demonstrated a failure rate of 80% with casting (n=10) and were better managed by closed reduction and percutaneous pinning (n=7, 100% success, P=0.002). Song 4 (n=325) fractures had low success rate (34%) with casting (n=35), but achieved higher success rates (P<0.001) when managed with either closed (n=57) or open reduction (n=233) and pin fixation (89.5% and 92.7% success, respectively, P=0.401). Song 5 fractures (n=149) generally required an open reduction in our series with good success rates (91.2%).
Conclusion: This study validates the Song classification with high interobserver and intraobserver reliability. The Song classification improves on existing classification systems by better distinguishing fractures at risk for failure of nonoperative treatment and guiding treatment outcomes.
Level of evidence: Level IV.