Discriminant analysis: a model for the prediction of relapse in Class III children treated orthodontically by a non-extraction technique.

Discriminant analysis has been used to predict the long-term outcome of treatment in children with Class III malocclusions considered suitable for orthodontic correction by a non-extraction technique. Thirty-four children, whose treatment included the application of headgear to the mandibular dentition, formed the basis of this study. Records were examined at the start of treatment and at least 2 years out of all retention. For 25 of these patients, the treatment outcome was unambiguous and from these data a four-variable discriminant model was generated using a step-wise selection procedure, run under SPSS. This afforded 100 per cent correct prediction of the relapse status of all 25 children. The model provided realistic predictions for the remaining nine cases and for seven out of eight other children considered suitable for treatment without extractions, but by a variety of different techniques. The model was not suitable for the prediction of treatment outcome in children who required extractions: these represented a distinctly different sub-group of Class III individuals. Thus, using four measurements taken from the start of treatment records, the eventual outcome of orthodontic treatment can reliably be estimated, provided that the model is restricted to the sub-group of Class III children for whom it was designed. If the discriminant model predicts that this form of therapy will relapse, an alternative approach may be chosen.