PREDICTION OF CLASS III TREATMENT NEED AND SUCCESS

Objective: The purpose of the present study is to develop prognostic models for surgical need and treatment success for class III malocclusions. Material and Methods: This is a retrospective cohort study that evaluated treatment outcomes of consecutively treated patients at UCSF from Jan 1st 2007-Jan 1st 2012 and UoP from May 1st, 2014 – May 1st 2019. Receiver operator curves were used to develop prognostic models for surgical need and treatment success for class III malocclusions. Predictor variables were selected a priori (Class III-WITS, U1-PP, IMPA). The prognostic models were validated first using a UCSF validation cohort to show consistency with in one program, and then using consecutively treated patients at UoP from May 1st, 2014 – May 1st 2019 as a second validation group as an outside program. Results: Derivation model for surgical need of class III malocclusion showed high sensitivity (81.8%); high specificity (94.3%), high positive predictive value (81.8%), high negative predictive value (94.3%), and the model correctly classified 91.3% of the subjects. UCSF validation model for surgical need of class III malocclusion showed moderate sensitivity (63.6%), high specificity (91.4%), high positive predictive value (70.0%), high negative predictive value (88.9%), and the model correctly classified 84.8% of the subjects. UoP validation model for surgical need of class III malocclusion showed moderate sensitivity (46.7%), high specificity (97.4%), high positive predictive value (77.8%), high negative predictive value (90.4%), and the model correctly classified 89.1% of the subjects. Derivation model for treatment success of Class III malocclusions showed moderate sensitivity (46.7%); high specificity (85.2%), moderate positive predictive value (63.6%), high negative predictive value (74.2%), and the model correctly classified 71.4% of the subjects. UCSF validation model for treatment success of Class III malocclusions showed low sensitivity (35.0%), moderate specificity (69.6%), moderate positive predictive value (50.0%), moderate negative predictive value (55.2%), and the model correctly classified 53.5% of the subjects. UoP validation model for treatment success of Class III malocclusions showed low sensitivity (16.1%), high specificity (87.5%), moderate positive predictive value (41.7%), moderate negative predictive value (65.3%), and the model correctly classified 62.1% of the subjects. Conclusion: WITS, U1-PP and IMPA were significant predictors of orthognathic surgical need in the derivation group, but only WITS predicted surgical need in the validation groups of Class III Malocclusions. Regarding treatment success, in the derivation group, only U1-PP was significantly associated with treatment success, while no variables were significantly associated with treatment success in the validation groups. Overall, the prognostic models developed in this study are more robust regarding predictions of Class III surgical need, as opposed to treatment success as defined by the ABO Cast and Radiograph examination