Methods: In order to accurately and consistently simulate the skull architecture, 7 skull phantoms were imaged using traditional cephalometry, CBCT, a high dose MSCT and a low dose MSCT. The data was imported into the Dolphin 3D software package (Dolphin Imaging and Management Systems, Chatsworth, CA, USA) and three dimensional models were constructed. These were used to synthesize 2D lateral cephalograms with a perspective projection intended to simulate the magnification of traditional 2D cephalometry. Three operators made 6 linear and 7 angular measurements in Dolphin, representing anteroposterior and vertical components of the craniofacial form. The measurements were made 3 times in random order, each time after a 1 week interval. The MANOVA test procedure was used to compare the different radiographic modalities for each measurement. Intraclass correlation coefficients (ICC) were used to determine intra- and inter-observer agreement for each cephalometric variable.
Results: High levels of inter- (ICC=0.92 to 0.99) and intra- (ICC=0.89 to 0.99) observer agreement were recorded. The linear measurements upper anterior face height (P=0.016), lower anterior face height (P=0.008) and posterior face height (P=0.008) and the angular measurements Sella-Nasion-B point (P=0.009) and the Frankfort mandibular plane angle (P=0.008) were statistically different between the modalities. Conclusions: Caution should be exercised when comparing measurements from 2D cephalograms synthesized from 3D scans, with conventional 2D cephalometric radiographs. These new simulated cephalograms may not be suitable for longitudinal research. Further research is required in this emerging field, before the clinical use of simulated 2D cephalograms can be recommended.