A Predictive Model for Assessing the Difficulty of Orthodontic Treatment of Impacted Maxillary Canines Using Cone-Beam Computed Tomography

Abstract

Introduction. Impaction of maxillary canines is one of the most common eruption anomalies (prevalence: 0.92–3.0%) and represents a significant clinical challenge due to the risks of root resorption of adjacent teeth, occlusal disturbances, and aesthetic defects. Despite the existence of various classification systems (e.g., KPG, TDI), most are purely descriptive and fail to provide a quantitative prognostic assessment of future treatment difficulty. Moreover, existing models largely ignore the craniofacial context — particularly the facial skeletal growth pattern and the condition of the upper airways, which may significantly affect the efficiency of orthodontic traction.

Aim. To develop and internally validate a multifactorial predictive model based on quantitative CBCT-derived parameters (local characteristics of the impacted canine and global craniofacial variables) for the objective assessment of orthodontic treatment difficulty for impacted maxillary canines in children and adolescents aged 10–14 years.

Materials and Methods. A retrospective cohort study was conducted involving 174 patients of the main sample (102 — LOI 0, 44 — LOI 1, 28 — LOI 2) and 47 healthy controls. The Level of Orthodontic Intervention (LOI) was determined post hoc from the medical records. All patients pre-treatment CBCT using an i-CAT Gendex CB-500 device. Seven craniofacial parameters (SNA, SNB, ANB, SN–GoGn, SN–PP, PNS–PPA, and nasopharyngeal volume) and seven local parameters of the impacted canine (angles β and α, vertical depth, shortest distance to the lateral incisor root, contact, follicular space width, and root dilaceration) were assessed. Measurements were performed by two blinded, calibrated examiners (ICC > 0.90). Ordinal logistic regression with stepwise selection, a multicollinearity check (using variance inflation factor, VIF), and the Brant test were used for model building. Internal validation was performed using 10‑fold cross‑validation.

Results. Four independent predictors entered the final model: vertical depth of the canine (OR = 2.85; 95% CI: 1.90–4.28; p < 0.001), shortest distance to the lateral incisor root (OR = 0.42; 95% CI: 0.28–0.63; p < 0.001), nasopharyngeal volume (OR = 0.55; 95% CI: 0.38–0.79; p = 0.001), and angle to the occlusal plane (OR = 1.78; 95% CI 1.12–2.83; p = 0.015). The model showed good explanatory power (Nagelkerke R² = 0.62) and statistical significance (χ² = 68.4; p < 0.001). A mathematical equation for the linear predictor Z and cut‑off values (τ₁ = 0.8; τ₂ = 2.1) were developed to stratify patients into three difficulty levels: LOI 0 (Z ≤ 0.8), LOI 1 (0.8 < Z ≤ 2.1), and LOI 2 (Z > 2.1).

Conclusions. The proposed predictive model allows for a quantitative assessment of orthodontic treatment difficulty for impacted maxillary canines prior to treatment initiation based on four key CBCT parameters. Although the model requires external validation on multicenter samples, it can currently serve as an auxiliary tool for patient stratification, treatment strategy optimization (observation, early surgical exposure, or autotransplantation), and for counseling patients regarding expected treatment duration and risks.