Reliability of end, neutral, and stable vertebrae identification in adolescent idiopathic scoliosis

Study Design. Analysis of radiographic interpretation and vertebral level identification. Objectives. To assess the intra- and interobserver reliability by observer training level used for selecting the end vertebra (EV), neutral vertebra (NV), and stable vertebra (SV) in adolescent idiopathic scoliosis patients. Summary of Background Data. Various radiographic and clinical factors are important in surgical planning. For adolescent idiopathic scoliosis, an analysis of the end, neutral, and stable vertebrae are of paramount importance for understanding spinal deformity management and determining the distal fusion level. Additionally, the development and comparison of optimal surgical techniques requires reliable, reproducible radiographic parameters. Methods. One hundred consecutive radiographs of operative cases of adolescent idiopathic scoliosis were evaluated on three separate occasions by three surgeons (2700 data points) at various levels of training (fellowship-trained spine surgeon, fellow in-training, orthopedic surgery resident). For each iteration, the observers attempted to identify the distal structural Cobb curve EV, NV, and SV. The radiographs included preselected Lenke type 1, 3, and 5 curves in random order. The average main thoracic curve was 53° (range, 30-82°) with a T8-T9 average apex, whereas the average thoracolumbar curve was 33° (range, 18-65°). Intra- and interobserver reliability was assessed by means of Cohen's Kappa correlation coefficient, and raw percentages of agreement were recorded. Results. Intraobserver reliability was good to excellent for determining the EV (κ_a = 0.69-0.88), good for determining the NV (κ_a = 0.65-0.73), and good to excellent fordetermining the SV (κ_a = 0.74-0.91) with 83.5, 72.2, and 85.6% intraobserver agreement, respectively. A trend was noted towards greater intraobserver reliability with increasing levels of observer experience. Interobserver reliability was poor (κ_a = 0.26-0.39) for each vertebral level, with interobserver agreement for only 48.7% of EV, 41.7% of NV, and 51.0% of SV. However, interobserver agreement increased significantly when concurrence within one vertebral level was assessed, with 91, 73, and 76% agreement for identifying the EV, NV, and SV, respectively. Conclusions. Radiographic determination of the EV, NV, and SV demonstrated good to excellent intraobserver, but poor interobserver, reliability. Interobserver agreement was fair to good when concurrence within one adjacent level was assessed. Observer experience level may be a factor. The difficulties in identifying these vertebral levels represent a potential obstacle to reproducible patient-specific fusion level determination and to the optimization and uniformity of patient care.