Retrospective clinical, radiographic, and patient outcome review of surgically treated adolescent idiopathic scoliosis. To correlate radiographic and clinical features of shoulder balance and the proximal thoracic curve with patient satisfaction outcomes at a minimum 2-year follow-up. Traditionally, radiographic features of a structural proximal thoracic curve have been T1 tilt, proximal thoracic Cobb angle, and proximal thoracic side-bending Cobb; however, these do not always correlate with clinical shoulder balance. A total of 112 patients (single surgeon) with adolescent idiopathic scoliosis and a proximal thoracic curve >or=20 degrees (average 32 degrees, range 20-78 degrees) were evaluated in terms of shoulder balance and curve flexibility/correction. Four groups were analyzed: Group 1, posterior spinal fusion to T2 (proximal thoracic curve included, n = 24); Group 2, posterior spinal fusion to T3 (proximal thoracic curve partially included, n = 23); Group 3, posterior spinal fusion to T4 or T5 (proximal thoracic curve not included, n = 21); and Group 4, anterior spinal fusion to T4 or below (proximal thoracic not included, n = 44). Proximal thoracic, main thoracic, and thoracolumbar-lumbar upright coronal, side-bending, and sagittal Cobb measurements were assessed before surgery, 1 week after surgery, and at a minimum 2-year postoperative follow-up (average 3.8 years, range 2.0-7.6 years). In addition to T1 tilt, clavicle angle (intersection of a horizontal line and the tangential line connecting the highest two points of each clavicle), coracoid height difference, trapezius length (horizontal distance of the T2 pedicle to second rib-clavicle intersection), first rib-clavicle height difference (vertical distance of first rib apex to superior clavicle), and proximal thoracic, main thoracic, and thoracolumbar-lumbar apical vertical translation were determined. Shoulder asymmetry as measured by the radiographic soft tissue shadow was graded as balanced (<1 cm), slight (1-2 cm), moderate (2-3 cm), or significant (>3 cm). A postoperative patient questionnaire addressed shoulder balance and overall appearance at most recent follow-up. The four groups were found to be statistically equivalent in terms of preoperative proximal thoracic curve (P = 0.4146), proximal thoracic side-bending Cobb (P = 0.2199), main thoracic curve (P = 0.6999), and main thoracic side-bending curves (P = 0.7307). Radiographic: Preoperative proximal thoracic measurements correlating with postoperative shoulder balance (P < 0.05) included the clavicle angle (three of four groups with a trend toward statistical significance in the fourth group, P = 0.07) and coracoid height (two of four groups). No other measurement, including T1 tilt and proximal thoracic side-bending Cobb, correlated in more than one group. Proximal thoracic curve correction was greatest in Group 1 (posterior spinal fusion to T2; average 12 degrees) and Group 4 (anterior spinal fusion to T4 or below; average 12 degrees). Clinical: Shoulder balance improved in all four groups (range 0.38-1.00 grades). There was no difference in shoulder balance between groups (P = 0.2723). Patient assessment: All four groups also reported improvement in self-perceived shoulder balance (63% up to one grade, 37% over two-grade improvement), whereas no patient reported worsening of shoulder balance. There was no significant difference in patient outcomes between the four groups (P = 0.3654). The clavicle angle, not T1 tilt, upright proximal thoracic, or side-bending proximal thoracic Cobb, provided the best preoperative radiographic prediction of postoperative shoulder balance. In each of the four groups, postoperative shoulder balance and clinical appearance also improved and correlated with patient postoperative assessments.