Table 5. Conclusions and limitations of the included studies.

Study	Conclusions	Limitations
Almukhtar et al., PLoS One (2014)	No differences between Voxel-based registration and Surface-based registration. High inconsistency between VBR and SBR regarding soft tissues.	I. No method error. II. In SBR, hard and soft tissues were superimposed separately whereas in VBR, hard and soft tissues were all superimposed at once.
Bazina et al., Am J Orthod Dentofacial Orthop (2018)	The Dolphin 3D software seems to work properly for voxel-based registration in the anterior cranial base.	I. The original change that occurred over time is not reported. II. ICC values were calculated from only 10 patients and for the average of all measurements. III. There was no assessment of the reproducibility of each individual measurement/case. IV. The type of ICC used is not reported.
Cevidanes et al., Dentomaxillofac Radiol (2005)	The technique shows acceptable reproducibility in the assessment of relatively unaltered structures.	I. There were relatively large interobserver errors compared to the detected changes. II. The actual measured changes were originally small (<0.8 mm).
Cevidanes et al., Am J Orthod Dentofacial Orthop (2009)	The technique provides reproducible 3D assessment of growing patients.	I. Small sample size that did not allow statistical comparisons.
Gkantidis et al., PLoS One (2015	Superimposition of 3D surface models created from voxel data can provide accurate, precise and reproducible results when appropriate references are used. Superimposition on BZ could be an alternative to AC.	I. CT data were used. II. No assessment of individual measurements regarding reproducibility.
Ghoneima et al., Orthod Craniofac Res (2017)	Surface-based and Voxel-based superimposition methods using the anterior cranial base as reference seem to be reproducible whereas Landmarks-based superimposition is less reproducible.	I. The original change that occurred over time is not reported. II. ICC values were calculated from only 10 patients and for the average of all measurements. There was no assessment of the reproducibility of each individual measurement/case. III. The type of ICC used is not reported. IV. The time span between serial images is not reported.
Häner et al., Orthod Craniofac Res (2019).	The Dolphin voxel-based superimposition technique exhibited adequate performance in growing patients, in terms of efficiency, cranial base matching, and reproducibility. The segmentation error was also acceptable in most cases.	I. The trueness of the voxel-based superimposition was assessed through visual inspection of corresponding reference structures in 2D. II. The original changes between T0 and T1 were relatively limited, though no relation was evident between the amount of change and the error of the process.
Koerich et al., Int J Oral Maxillofac Surg (2016)	The technique shows high precision and reproducibility tough these were assessed in relatively unaltered structures. Furthermore, differences between reoriented dry skulls were larger than expected.	I. The changes of the structures that were evaluated were quite small (<0.3 mm). II. Differences between the serial images of reoriented dry skulls were higher than those of actual serial scans. III. Samples from different machines were tested.
Koerich et al., Angle Orthod (2017)	The technique shows moderate reproducibility in the assessment of relatively unaltered structures.	I. Relatively large interobserver errors compared to the detected changes. II. The changes measured were small (<0.9 mm).
Lemieux et al., Am J Orthod Dentofacial Orthop (2014)	The landmark-derived maxillary plane cannot be assessed through the present methodology.	I. The main outcome is not suitable for the assessment of the superimposition result because it remains unaffected by the superimposition itself. II. The landmark identification error is not thoroughly assessed for individual cases.
Nada et al., PLoS One (2011)	This technique might show good trueness and reproducibility. Registration on the left zygomatic arch seems to be less accurate, but it might still be clinically acceptable and reproducible.	I. Only structures considered stable were evaluated and thus the measured changes were small. II. Only mean values are provided and analysed and thus possible larger individual differences are ignored.
Nguyen et al., Am J Orthod Dentofacial Orthop (2017)	The chin and the symphysis region might be an anatomically stable reference area for mandibular superimpositions, whereas the third molar region displayed a higher instability. The chin+symphysis area seems to provide reproducible results.	I. The bone plates and screws were confirmed to be immobile clinically, but their stability in space was not tested (e.g., through best fit registration). II. The areas identified as stable were located at the same place where the superimposition reference area was. III. Only average measures were used to assess all outcomes. There was no assessment of individual cases. IV. Reproducibility outcomes were tested assessing the whole mandibular surface. V. The performance of the chin+symphysis area was shown only for 1 subject.
Ruellas et al., Am J Orthod Dentofacial Orthop (2016a)	No clear evidence is provided that the 2 regions of maxillary registration show similar results and adequate intraobserver and interobserver reproducibility values for growing patients.	I. The changes measured were originally small, except from landmarks 2 and 6 where the error was greater. II. In individual cases the amount of differences was not small compared to the original changes. III. No detailed information is provided (e.g., Bland Altman plot for every variable and results for any landmarks and every coordinates). Only means of different variables were assessed. IV. No comparative statistics.
Ruellas et al., PLoS One (2016b)	The body of the mandible might show better agreement with direct measurements from a point considered stable, compared to the modified Björk superimposition.	I. Results from one superimposition technique (Björk) are not reported. II. The gold standard reference values were not reliable since one geometrical point that was speculated to be stable was used to generate them. However, landmark point identification error is expected to be high in this case and this was not evaluated. III. Two cases were not included in the analysis. IV. Reported mean values were outside of the Limits of Agreements.
Weissheimer et al., Int J Oral Maxillofac Surg (2015)	The software seems to be user-friendly and might work properly for voxel-based registration in the anterior cranail base, both for growing and non-growing patients.	I. No results quantification. II. No method error. III. No descriptive and comparative statistics. IV. Only data from 2 patients shown.