Early prediction of maxillary canine impaction

Objectives: The aim of this study was to establish prediction criteria for maxillary canine impaction in young patients, based on angular and linear measurements on panoramic radiographs. Methods: From 828 records having at least 2 panoramic radiographs, both taken between the ages of 7 and 14 years, with a minimum 1-year and maximum 3-year interval (T1 and T2), a training data set consisting of 30 subjects with unilateral canine impaction (12 males and 18 females) was selected. The patients' mean age was 10.1 years standard deviation (SD) 1.3 years at T1 and 11.9 years (SD 1.1 years) at T2. The training data set also consisted of 30 maxillary canines from the contralateral sides and an additional 60 normal erupted canines from 30 subjects. Those 30 subjects of a test data set were selected based on displaying bilateral maxillary canine eruption at T2 and being matched for gender and age with the subjects of the training data set 12 males and 18 females; mean age at T1, 10.1 years (SD 1.3 years) and at T2, 11.1 years (SD 1.2 years). Angular and linear measurements were performed separately by two observers on the total study sample at T1. Linear measurements were expressed as a multiplication of the maxillary central incisor width at the non-impacted side. Results: Significant differences for linear and angular measurements and radiographic factors were found between the maxillary impacted canine and erupted maxillary canine. The three best-discriminating parameters were canine to first premolar angle, canine cusp to midline distance and canine cusp to maxillary plane distance. These three parameters were combined in a multiple logistic regression model to calculate the probability of impaction, yielding a high area under the curve (AUC) equal to 0.97 (95% confidence interval: 0.94–0.99), with 90% sensitivity and 94% specificity. Conclusions: Prediction of maxillary canine impaction from a combination of parameters relating to angles and distances measured in panoramic radiographs is weak. However, the final prediction model, based on canine–first premolar angle, canine cusp tip to midline distance and canine cusp tip to maxillary occlusal plane distance, might be useful to discriminate canine impaction for early intervention or regular follow-up.