Prognosis versus actual outcome. III. The effectiveness of clinical parameters in accurately predicting tooth survival.

Tooth loss for 100 treated periodontal patients (2,509 teeth) under maintenance care was evaluated to determine the effectiveness of commonly taught clinical parameters utilized in the assignment of prognosis in accurately predicting tooth survival. Previous studies in this series evaluated prognosis as a surrogate variable representing the condition of the tooth at a particular point. In this study, survival analysis was used to evaluate the relationship of these common clinical parameters to an actual end point, tooth loss. Robust log rank tests indicated that initial probing depth, initial furcation involvement, initial mobility, initial crown-to-root ratio, and initial root form were all associated with tooth loss. In addition, smoking and increased initial bone loss were both found to be associated with increased risk of tooth loss while fixed abutment status was associated with a decreased risk of tooth loss. A Cox proportional hazards regression model showed that initial probing depth, initial furcation involvement, initial mobility, initial percent bone loss, presence of a parafunctional habit without a biteguard, and smoking were all associated with an increased risk of tooth loss. This model suggests that patients are twice as likely to loose their teeth if there is increasing mobility, if they have a parafunctional habit and do not wear a biteguard, or if they smoke. From these data there does appear to be a relationship between the assigned prognosis and tooth loss. Teeth with worse prognosis have a worse survival rate, but the commonly taught clinical parameters used in the traditional method of assignment of prognosis do not adequately explain that relationship. Furthermore, initial prognosis did not adequately explain the condition of the tooth or accurately predict the tooth's survival. These results seem to indicate that the effect of these clinical parameters on tooth survival is only partially reflected in the assigned prognosis initially, suggesting that perhaps some of the clinical parameters should be weighed more heavily than others when assigning prognosis. Further studies are needed to develop a more accurate method for the assignment of prognosis.