Multifactorial modeling of temporomandibular anatomic and orthopedic relationships in normal versus undifferentiated disk displacement joints

Statement of problem. There is persistent dispute about the diagnostic value of hard tissue anatomic relationships in predicting temporomandibular joint disorders and normals. Purpose. The goal of this study was identification of multifactorial temporomandibular hard tissue relationships that differentiate asymptomatic normal joints. Material and methods. Central section lateral tomograms of 162 female temporomandibular joints with pooled diagnoses of unilateral disk displacement with and without reduction were compared to 42 female asymptomatic normal joints using 14 linear and angular measurements and 8 ratios. A validated classification tree model was tested for accuracy with sensitivity, specificity, goodness of fit, and the amount of the log likelihood accounted for. The tree model was compared with a multiple logistic regression model and univariate testing. Results. The classification tree model consisted of 3 asymptomatic and 4 disk displacement terminal nodes consisting of interactions of condyle position with measures of fossa size and shape, of which mainly average nonextreme measurements and more frequent concentric ranges typified the asymptomatic joints. The logistic regression and univariate models also incorporated condyle position and size, but the logistic regression accounted for less of the log likelihood than the tree (23.3% vs. 32.6% Rescaled Cox and Snell R²). The tree and the logistic regression models were moderately good predictors for distinguishing normals from disk displacement joints (sensitivity 67.9% and 72.2%, specificity 85.7% and 76.2%, respectively). Although the univariate analysis showed that the asymptomatic joints had smaller mean fossa width to fossa depth ratios (P<.0005), shorter mean eminence length (P<.007), and more concentric to anterior mean condyle position (P<.049), overlap in most of the ranges limited the predictive value. Conclusion. Within the limitations of this study, multifactorial analysis revealed that several subsets of asymptomatic temporomandibular joints could be distinguished from joints with disk displacement according to hard tissue measurements taken from central section tomograms. In general, asymptomatic normal joints were typified by interactions of less extreme ranges of fossa size, shape, and condyle position.