Stepwise Regression Model in Predicting Excessive Mandibular Growth

Predicting excessive mandibular growth in developing Class III patients remains a challenge to clinicians. Objective: To determine the cephalometric measurements that can be used to predict excessive growth in Class III patients using longitudinal treatment records. Methods: Lateral cephalometric radiographs were collected from thirty patients with skeletal Class III malocclusion. All patients were previously treated with protraction facemask and followed for a minimum of four years. Patients were divided into two groups: Patients maintaining a positive overjet following facemask treatment were classified as successful (n=15). Patients relapsing back into an anterior crossbite and requiring orthognathic surgery were classified as unsuccessful (n= 15). Lateral cephalometric radiographs before and after treatment were traced. Thirteen cephalometric variables (AB to mandibular plane angle, Wits, ANB, Go angle, Ar-Go, Ar-Me, PTV-A, PTV-A/Ar-Me, N-ANS, ANS-Me, N-Me, ANS-Me/N-ANS, Po-GD) were measured and recorded. Data was analyzed using stepwise logistic regression, discriminant analysis, and regression analysis. Results: Stepwise variable selection showed that mandibular length (Ar-Me) was able to accurately predict 86.7% of the successful cases and 61.5% of the unsuccessful cases. When patients were followed over a period of time, significant difference in growth rate and direction was found with ramal length (Ar-Go) and gonial angle (Ar-Goi-Me). Gonial angle and ramal length were found to increase over time with the unsuccessful group. Conclusions: These results suggest that the cephalometric variables mandibular length (Ar-Me), ramal length (Ar-Go), and gonial angle (Ar-Goi-Me) can be used in predicting successful and unsuccessful outcome in early Class III orthopedic treatment.