Study design: Statistical analysis of various measurement techniques for thoracolumbar burst fracture kyphosis on lateral radiograph.
Objective: To determine the most reliable measurement technique.
Summary of background data: The treatment of thoracic and lumbar burst fractures involves many factors, including the degree of resultant kyphosis. Although various methods have been described, no study has directly compared these methods for reliability and reproducibility.
Methods: Fifty lateral radiographs of thoracic and lumbar burst fractures were randomly selected and measured on two separate occasions by three spine surgeons using five different measurement techniques. Radiograph quality, fracture type, and the center beam location were determined. Statistical analysis included analysis of variance for repeated measures and analysis of variance using a generalized linear model.
Results: Intraclass correlation coefficients were most consistent for Method 1 (rho = 0.83-0.94) followed by Method 4 (rho = 0.65-0.89) and Method 5 (rho = 0.73-0. 85). Intraobserver agreement (% of repeated measures within 5 degrees of the original measurement) ranged between 72% and 98% for all techniques for all three observers, with Method 1 showing the best agreement (84%-98%). Paired comparisons between observers varied considerably with interobserver reliability correlation coefficients ranging from 0.52 to 0.93. Method 1 showed the highest interobserver reliability coefficient (0.81, range 0.71-0.93) followed by Method 5 (0.71, range 0.68-0.75). Method 1 also had the highest percentage of agreement within categories (90% within 5 degrees ).
Conclusions: Method 1 (measuring from the superior endplate of the vertebral body one level above the injured vertebral body to the inferior endplate of the vertebral body one level below) showed the best intraobserver and interobserver reliability overall.