A prospective study comparing supine bending, push-prone, and traction under general anesthesia (UGA) radiographs in adolescent idiopathic scoliosis. To compare the effectiveness of 3 different flexibility methods on structural and nonstructural main thoracic (MT) and thoracolumbar/lumbar (TL/L) curves. Supine side bending radiographs are the major method for determining curve reducibility. Push-prone radiographs show structural and compensatory curves on the same radiograph, but have not shown comparative reducibility in recent studies. Traction UGA is a relatively new modality which may offer similar or improved flexibility, while also showing structural and compensatory curves on the same radiograph. Fifty-eight patients with adolescent idiopathic scoliosis were prospectively studied with standing PA and lateral, supine bending and push-prone radiographs before surgery; traction UGA radiographs intraoperatively; and standing PA and lateral radiographs after surgery. Traction UGA demonstrated equal flexibility to supine bending in structural MT and TL/L curves, with a trend towards more flexibility in severe MT curves > or = 60 degrees, while push-prone demonstrated significantly less flexibility. Each of the methods showed significantly less flexibility than postoperative correction in MT curves, while traction UGA and supine bending were not significantly different than postoperative correction in TL/L curves. Analysis of nonstructural TL/L curves showed that traction UGA and push-prone were not significantly different than postoperative correction, while supine bending overestimated postoperative correction. Traction UGA offers flexibility equivalent to supine bending for structural MT and TL/L curves, and flexibility comparable with push-prone for nonstructural TL/L curves. Traction UGA also shows both structural and compensatory curves on the same radiograph, and ultimately may provide a better estimate of spinal balance.