Fig 1.
CONSORT flow diagram of participant enrollment, randomization, allocation, follow-up, and analysis.
A total of 162 individuals were screened for eligibility. Twenty-two were excluded (15 not meeting inclusion criteria, 5 declined to participate, and 2 for other reasons). The remaining 140 were randomized equally to the Digital and AI group (n = 70) or the Conventional group (n = 70). All participants completed the allocated intervention, attended follow-up visits at T1-6m, T1-12m, and T2, and were included in the final analyses.
Table 1.
Baseline demographic and clinical characteristics of the Digital and AI group and the Conventional group.
Table 2.
Total PAR scores at baseline (T0), 6 months (T1-6m), 12 months (T1-12m), and treatment completion (T2).
Table 3.
Proportion of participants achieving different levels of PAR reduction at T1-6m, T1-12m, and treatment completion (T2).
Fig 2.
Categorical and temporal distributions of PAR improvement.
(A) Stacked bar charts showing the proportions of participants achieving significant improvement (≥70%), partial improvement (30–70%), and no improvement (<30%) in the Digital and AI group and the Conventional group at T1-6m, T1-12m, and T2. (B) Time trend of the proportion of participants achieving significant improvement (≥70%) in PAR scores from baseline (T0) to T2. The Digital and AI group consistently showed higher rates than the Conventional group. Error bars represent 95% confidence intervals, and p values indicate between-group differences at each timepoint. This figure reflects differences in occlusal improvement only and should not be interpreted as evidence of enhanced treatment efficiency, reduced chairside time, or accelerated workflow, as such metrics were not collected in this study. Repeated-measures mixed-effects analysis demonstrated significant intervention (p < 0.001), time (p < 0.001), and group × time interaction effects (p < 0.001).
Table 4.
PAR component scores and total PAR score at baseline (T0), 6 months (T1-6m), 12 months (T1-12m), and treatment completion (T2).
Fig 3.
Dynamic changes in PAR scores across treatment stages in the DAOG and COG.
(A) Upper anterior; (B) Lower anterior; (C) Overbite/Open bite; (D) Overjet; (E) Midline alignment; (F) Buccal occlusion. Data are presented as mean ± SD. Significant differences between groups at T2 are marked in each panel.
Table 5.
Subgroup analysis of baseline PAR, post-treatment PAR, and improvement rates by age category.
Fig 4.
PAR improvement rates (%) at T2 by age subgroup.
Violin plots show the distribution of PAR improvement rates (%) at T2 for age subgroups (12–18, 19–25, 26–35 years). Each dot represents an individual participant. Box plots indicate interquartile ranges with medians, and black diamonds represent means with 95% confidence intervals. Horizontal lines and p values indicate between-group differences (independent-samples t-test). Subgroup analyses were exploratory and underpowered; therefore, non-significant interaction results should not be interpreted as evidence of uniform treatment effects across age groups.
Table 6.
Multivariable regression analysis of factors associated with PAR score improvement.
Fig 5.
Forest plot of multivariable regression predictors of PAR improvement.
Data are shown as β-coefficients with 95% confidence intervals (CI). Predictors included treatment modality, baseline PAR score, age group, and gender. Significant predictors are indicated when CIs do not cross zero.
Table 7.
Final PAR scores at T2 after exclusion of outliers (sensitivity analysis).
Fig 6.
Sensitivity analysis of PAR improvement rates before and after outlier exclusion.
Paired boxplots illustrating PAR improvement rates (%) before and after outlier exclusion in the Digital and AI group and the Conventional group. Each dot represents an individual participant. Lines represent paired observations. Statistical significance of within-group differences was determined using paired t-tests (**p < 0.01, **p < 0.001).