[Impact of different registration methods on the accuracy of virtual occlusal records in implant restoration for multiple missing teeth: an <i>in vitro</i> study].

<b>Objective:</b> To investigate the impact of four registration methods on the accuracy of virtual occlusal records (VOR) in intraoral scanning for implant restorations with multiple missing teeth. <b>Methods:</b> A mandibular model simulating clinical conditions with multiple missing teeth (right first molar, left second premolar, left first molar, left second molar) and a maxillary compete dentition model were mounted on a semi-adjustable articulator. Subsequently, twelve 0.5 mm stainless steel spheres were adhered to reference positions (16, 46, 13, 43, 23, 33, 25, 35, 26, 36, 27, 37) as fiducial markers. Following this, a laboratory scanner generated reference datasets by digitizing the models in maximum intercuspation (MIP). Meanwhile, ten maxillary and mandibular scans were acquired using an intraoral scanner, with all nonarticulated scans duplicated four times to ensure data consistency. Forty VOR intraoral scans were performed in MIP using four registration protocols: left-side, right-side, anterior, and bilateral registration (<i>n</i>=10 per group), randomized via a computer-generated pseudo-random sequence. For measurement, linear distances (D16-46, D13-43, D23-33, D25-35, D26-36, D27-37, D16-46 represented the single-tooth defect position, whereas D25-35, D26-36, D27-37 reflected positions in free-end edentulism areas) between opposing markers were measured in a reverse engineering software, with deviations (ΔD) from the reference scan calculated to assess accuracy. Specifically, negative ΔD values indicated vertical dimension underestimation. Given that non-normally distributed data were analyzed using medians [interquartile ranges (IQR)], trueness (median ΔD) and precision (IQR) were evaluated. The interaction effect between the registration method and the position of the measurement items was evaluated by using the generalized linear model. The accuracy was compared overall by the Kruskal-Wallis test with the two-sided significance level of α=0.05. For pairwise comparisons, post-hoc tests were conducted by Dunn's <i>t</i>-test with the Bonferroni correction for the significance level. <b>Results:</b> The accuracy of VOR was affected by registration method (<i>P</i><0.05), with a significant position×registration method interaction observed (<i>P</i><0.05). In particular, in all four groups, only the bilateral registration group showed trueness of less than 0.1 mm for both free-end edentulism and the single tooth defect, with ΔD16-46, ΔD25-35, ΔD26-36, and ΔD27-37 being 0.059 (0.015), -0.082 (0.052), -0.065 (0.032), -0.070 (0.050) mm, respectively. Moreover, trueness in free-end edentulism showed negative values across all groups, with the largest negative deviations observed in the right-side registration group, with ΔD25-35, ΔD26-36 and ΔD27-37 being -0.410 (0.174), -0.442 (0.225), -0.439 (0.262) mm, respectively. <b>Conclusions:</b> In fully digital workflows of implant restorations for mandibular free-end edentulism with multiple missing teeth, registration method critically influences VOR accuracy. While four registration methods exhibited underestimation of occlusal vertical dimension, bilateral registration achieved the highest accuracy.