Table 2.

Summary of the articles included in the literature review (AlAli et al., 2017 [28], Chen et al., 2017 [5], Fasel et al., 2016 [25], Garas et al., 2018 [22], Jones et al., 2016 [36], Lee et al., 2018 [29], Li et al., 2017 [4], Lim et al., 2016 [23], Lozano et al., 2017, Mogali et al., 2018 [24], Smith et al., 2018 [21], Wu et al., 2018 [6], Young et al., 2018 [27])

Study	n	Geographic region	Year	Key points (“N” or “A” denotes whether normal or abnormal/uncommon anatomy was demonstrated)	Strengths	Weaknesses
RCT
Lim et al.	52	Australia	2016	Comparing the use of cadaveric materials/3D model/combined materials in teaching cardiac anatomy, the 3D model group performed best (p = 0.012) (N)	Randomised groups Assessor blinded to teaching	Difficult to explain why the “combined” group did not confer the same advantage
Chen et al.	79	China	2017	Coloured skull models provided better educational outcome than cadaveric skulls and atlas groups (p < 0.05), as evaluated with junior medical students (N)	Post-tests evaluated by both written and lab tests Inexpensive and precise skull models used	Lack of pre-tests Not blinded Defects in cadavers
AlAli et al.	67	UK	2017	Students were randomised into 2 groups (Powerpoint presentation vs Powerpoint +3D models) in the education of cleft lips and palates Post-tests showed improvement in knowledge with 3D model group (p = 0.038) (A)	Clear learning objectives Low cost	A heterogenous sample group (sampled from two universities) Not compared to cadavers (gold standard)
Smith et al.	127	UK	2018	Randomised groups (2D teaching vs 3D model) participated in anatomy training with a range of 3D models—3D group performed better Questionnaires also showed excellent feedback (N)	Mixed methods (also uses student focus group and questionnaires, faculty evaluation) Pre-tests showing same baseline between randomised groups A range of anatomical models used	Lack of intravascular contrast in a cadaver specimen meant that the vascular structures were indistinguishable from surrounding soft tissues of similar density, affecting fidelity Formats of pre- and post-tests were different but rationale was not explained (short answer vs single best answer, respectively)
Wu et al.	90	China	2018	Models with spinal, pelvic, upper limb and lower limb fractures were evaluated by medical students (radiographic image group vs 3D-printed model group)—3D model group performed better with the pelvic and spinal test (A)	Involvement of pathologies (fractures) Diverse range of bony anatomy	No improvement in the upper and lower limb tests—perhaps advantage was only applicable to more complex anatomy
Sequential comparison study
Lee et al.	20	Korea	2018	Renal models from ten patients with renal tumours were evaluated by urologists and medical students using questionnaires Students also located tumours more accurately using the 3D models compared to CT images alone (A)	Introduction of pathology into the models helped to mimic real-life clinical scenarios High-fidelity models Also taught on radiological anatomy	Not randomised Small sample size
Garas et al.	23	Australia	2018	Students were all exposed to 3D, wet and plastinated specimens of the heart, shoulder and thigh. With each model, they were asked to identify pinned structures with the aid of 2D atlases A larger number of participants achieved right answers for 3D models compared to wet and plastinated materials (N)	Used both wet and plastinated materials for comparison Covered three different anatomical parts	No pre-tests to assess baseline knowledge Not randomised Small sample size Contained a heterogenous mixture of samples (1st and 3rd year students)
Survey
Jones et al.	51	USA	2016	High-fidelity replicas of a number of surgical models were presented at a conference and feedback was provided from written survey (A)	A range of models evaluated (breast, lung, liver, aorta) High fidelity	Surveys were filled in by a mixture of surgeons (majority) and medical students
Mogali et al.	15	Singapore	2018	Multi-coloured and multi-material 3D models of the upper limb were compared with plastinated prosections using surveys and focus group discussion Anatomical features in 3D models were rated as accurate by all students (N)	High fidelity—uses multiple material	Compared to plastinated prosections but not soft-prepared cadavers Small sample size
Review
Li et al.	N/A	Hong Kong	2017	The authors provided a broad overview of applications of 3D printing in surgery (A)	Put the application of 3D models into a broader context Also reviewed several cardinal papers (some included here)	Focused on surgical education—especially postgraduate level
Others
Fasel et al.	12	Switzerland	2016	Anatomy teaching was undertaken by making measurements on replica, scans and cadavers Measurements from 3D-printed models were close to authentic anatomic reality (N)	Kinaesthetic approach to teaching by making measurements Also taught on radiological anatomy	Lack of a control group Likely more time-consuming Small sample size
Lozano et al.	N/A	Spain	2017	A process of manufacturing of a 3D-printed skull model was described (N)	Detailed description of the manufacturing process including both the soft and hardware settings	Did not specify how students were systematically involved in the evaluation of models
Young et al.	N/A	Australia	2018	CT imaging was used to create replicates of embryonic and fetal anatomical specimens, with good responses from students (A)	A novel concept that studied a difficult aspect of anatomical sciences Also lessened the potential for adverse student reaction (due to cultural background or personal experience)	Did not specify how students were systematically involved in the evaluation of models No quantitative results

RCT randomised controlled trial, CT computed tomography