This study aimed to compare the accuracy of VSP and CSP in bimaxillary orthognathic surgery, focusing on differences in positional accuracy across the X (medial–lateral), Y (anterior–posterior), and Z (superior-inferior) axes. The comparison of VSP and CSP revealed that the clinically achieved predictability of both methods was similar.

VSP and CSP demonstrated no significant differences between the planned and postoperative outcomes across all dimensions. When comparing VSP predictions with actual postoperative outcomes, no statistically significant differences were observed across all three axes. However, assessment of absolute discrepancies revealed that VSP exhibited greater deviations than CSP in certain maxillary landmarks, particularly in the anterior–posterior direction (Y-axis) at U3L, U1L, U1R, and U1 (p < 0.05), where VSP showed larger errors compared to CSP.

Our results demonstrated that 3D planning significantly differed from 2D planning at key maxillary landmarks, particularly in the Y-axis (anterior–posterior direction). Notably, U1L, U1R, U3L, U3R, and U6R showed significant differences indicating that VSP provided a more refined and accurate surgical plan compared to CSP. These findings are consistent with Ho [8], who also reported discrepancies in landmark positioning along the Y-axis. A possible explanation for this difference is that 3D planning allows for better visualization of yaw rotation, which are often not as evident in 2D planning methods. As a result, yaw adjustments are systematically incorporated in VSP to achieve optimal maxillary and mandibular symmetry. This continuous refinement in 3D surgical planning could contribute to the observed differences in anterior–posterior positioning of maxillary landmarks compared to CSP.

Our findings demonstrated that postoperative outcomes did not significantly differ from the 3D surgical plan, regardless of whether the surgical splints were fabricated using manual (CSP) or 3D-printed (VSP) techniques. This result highlights that the 3D surgical plan itself is the key determinant of surgical accuracy. 3D virtual planning provided a clear visualization of key surgical parameters, including bone contact points, bony gaps, and precise skeletal movements on the simulated surgical model. This enhanced preoperative understanding, combined with the use of surgical splints, enabled accurate replication of the digital treatment plan intraoperatively. The integration of both surgical guides splint and detailed 3D visualization ultimately facilitated precise execution of the surgical plan, contributing to optimal patient outcomes.

Comparison of the discrepancies between VSP and CSP relative to the planned surgical outcomes revealed no statistically significant differences between the two groups. This suggests that both VSP and CSP offer comparable accuracy in terms of overall surgical outcomes. However, analysis of absolute values of discrepancies revealed the VSP group exhibited larger deviations than the CSP group, particularly at U3L, U1L, U1R, and U1. One possible explanation for this discrepancy could be the difference in splint thickness and material properties between the two methods [9]. The VSP splints were thinner and had lower flexural strength compared to the CSP splints, which may have affected their resistance to deformation during surgery. Due to the manufacturing process, CSP splints tend to be thicker, which may have improved their stability intraoperatively [7, 9]. However, if the surgical splint is too thick, especially when the depth of the cusps embedded in the splint exceeds 3 mm, it may reduce surgical accuracy due to premature contacts between the teeth and the splint [17].

In our study, most patients presented with skeletal Class III deformities requiring anterior and inferior maxillary repositioning. Such movements often result in occlusal overlaps between the upper and lower jaws, requiring a vertical mandibular opening during virtual surgical planning to fabricate a viable intermediate splint. Consequently, the determination of a condylar hinge axis becomes a critical step. In VSP, this axis is commonly defined at the posterosuperior point of the condyle. However, existing literature reveals considerable interindividual variability in the instantaneous center of rotation (ICR), which shifts dynamically during mandibular movement. Previous studies have shown that errors in defining this rotational axis can lead to significant discrepancies in maxillary positioning, especially in the sagittal plane, due to inaccurate simulation of mandibular autorotation [6]. Therefore, precise hinge axis determination remains a critical focus for improving the accuracy of virtual planning in orthognathic surgery.

Another contributing factor may be inaccuracies in centric relation (CR) registration. Our protocol used wax bite records to define CR for 3D simulation. However, the concept of CR remains controversial. While the Glossary of Prosthodontic Terms defines it as the most anterior-superior condylar position, orthognathic surgeons often adopt a posterior-superior manipulation, termeds retruded contact position. Misalignment between clinical practice and theoretical definitions may result in unstable mandibular positioning, particularly in patients with complex occlusal discrepancies. This introduces spatial errors that cascade into maxillary misalignment during intermediate splint design and surgical execution [3].

Despite technological advances in 3D printing, reliance on splint-only transfer remains a limiting factor. A recent study reported vertical errors up to 5 mm in anterior maxillary positioning when using 3D-printed splints alone, highlighting the impact of surgical technique, lack of rigid reference points, condylar positioning variability, and bony interference during osteotomy [11].

Surgical sequencing may impact accuracy. While patient-specific osteosynthesis (PSO) systems typically offer higher precision, evidence shows that maxilla-first sequencing results in greater deviations compared to mandible-first protocols (1.8 mm vs. 1.0 mm; p = 0.008) and 40.5% of the cases had a deviation of > 2 mm in any direction at the upper incisor point, due to increased reliance on stable condylar seating during maxillary repositioning [16]. As our study relied solely on occlusal splints and manually bent fixation plates—without PSO—the likelihood of positional inaccuracies may be inherently higher.

Moreover, recent comparative studies highlighte the limitations of occlusal splints, whether manually fabricated conventional resin occlusal splint (CROS) or 3D-printed digital occlusal splint (DOS) [5]. Quantitative analysis of maxillary repositioning accuracy revealed mean deviations of 2.55 ± 0.95 mm for CROS, 2.15 ± 1.12 mm for DOS, and a significantly lower at 1.17 ± 0.66 mm for the digital template group (P < 0.001 vs. CROS; P = 0.001 vs. DOS). These findings highlight that, despite improvements in fabrication, occlusal splints—whether digital or conventional—remain dependent on mandibular positioning and lack vertical control of the maxilla. Notably, no significant difference was found between CROS and DOS, confirming that the core limitation lies in the splint-based technique itself rather than the fabrication method. In contrast, digital templates allowed maxillary positioning independent of mandibular positioning, leading to superior accuracy.

Further research is needed to assess the impact of splint thickness, condylar rotation modeling, and material properties on surgical accuracy. Additionally, splint-less approaches using customized titanium plates and cutting guides could serve as an alternative to reduce these discrepancies and improve vertical control.

While VSP improves preoperative visualization and surgical planning, critical technical factors—including hinge axis definition, CR registration, splint stability, and surgical sequence—continue to influence the final surgical outcome. To improve accuracy, future developments in VSP should prioritize dynamic mandibular modeling, standardized CR determination methods, and transition toward template-based or splint-less protocols for improved intraoperative control and reproducibility.

Comparison of VSP and model surgery, demonstrated no statistically significant differences for most maxillary landmarks, indicating that both methods provide comparable predictive accuracy. However, a trend toward under-correction in model surgery was observed at U6L in the Y-axis and Z-axis. Song et al. [13] also reported differences between cast model surgery and 3D planning in the Y-axis and Z-axis at the maxillary first molar position. These findings suggest that manual model surgery may introduce minor inaccuracies in vertical and anterior–posterior positioning.

Our results highlight the superiority of VSP over CSP for preoperative planning due to its improved precision and ability to simulate complex surgical movements in three dimensions. However, despite its advantages, VSP is not entirely error-free, particularly in anterior–posterior positioning. These discrepancies could be attributed to differences in intraoperative execution, soft tissue influences, and the complexity of translating digital plans into surgical practice.

The findings also emphasize that while VSP and model surgery yield comparable results, minor deviations in model surgery may still impact final outcomes, particularly in vertical and anterior–posterior positioning.

This study has several limitations. Most notably, the relatively small sample size limits the statistical power and generalizability of the findings. While the randomized controlled design strengthens the internal validity, the number of cases remains insufficient to draw definitive clinical conclusions. As such, this study serves as a pilot investigation that provides foundational data for larger-scale trials. Additionally, the analysis was limited to immediate postoperative outcomes, without evaluation of long-term skeletal stability. Future research should incorporate larger cohorts and longitudinal follow-up to comprehensively assess the clinical efficacy and durability of both VSP and CSP approaches.