BACKGROUND What is thought of as a classic "cubitus varus" deformity usually consists of varus, extension, and internal rotation. However, its 3-dimensional (3D) pattern with 3D imaging has not been reported. This study aimed to obtain such 3D patterns using 3D bone models created from computed tomography data and evaluate the accuracy of conventional radiographic and clinical methods of assessing the deformity. METHODS Imaging of 25 humeri of 25 patients with cubitus varus deformity caused by previous humeral supracondylar fractures was performed. The deformity was assessed by superimposing the 3D bone model onto a mirror-image model of the contralateral normal humerus. The 3D deformity pattern of cubitus varus was evaluated based on the 3 deformity components. Values obtained from conventional radiographic and physical measurements--that is, humerus-elbow-wrist angle (HEW-A), tilting angle (TA), maximal elbow flexion angle (MEF), and internal rotation angle (IRA)--were compared with those from the 3D technique. RESULTS Of the patients, 44% had varus, extension, and rotation deformities of 10° or greater; 20% had varus and extension deformities of 10° or greater; 16% had varus and internal rotation deformities of 10° or greater; and 20% had varus deformity only. When the 3D measurements were considered accurate, an error of 10° or greater was found in 8%, 24%, 8%, and 44% of cases in terms of HEW-A, TA, MEF, and IRA values, respectively. CONCLUSIONS Of the humeri, 80% had other bony deformities in addition to varus and 20% had isolated varus deformities. HEW-A and MEF showed reasonable accuracy as measures for the degree of deformity, whereas TA and IRA were found to be relatively inaccurate.