Multifactorial analysis of differences in temporomandibular joint hard tissue anatomic relationships between disk...

Without multifactorial models, it is difficult to resolve whether hard tissue tomographic relationships can distinguish differences between temporomandibular joint (TMJ) internal derangement diagnoses.

The purpose of this study was to use multifactorial models to examine whether there are hard tissue anatomic and orthopedic characteristics that distinguish temporomandibular joints with disk displacement with reduction from disk displacement without reduction.

. TMJ tomograms from female patients who had unilateral disk displacement diagnosed with (n = 84) or without (n = 78) reduction were compared with the use of 14 linear and angular measurements and 8 ratios. A representative classification tree model was tested for fit with sensitivity, specificity, accuracy, and likelihood accountability, and the results were compared with a multiple stepwise logistic regression model and univariate analysis.

Disk displacement without reduction joints had longer mean postglenoid fossa heights (P<.0005), greater mean fossa depth (P<.017), and narrower mean absolute superior joint spaces (P<.041) than disk displacement with reduction joints (univariate t test). The classification tree had 4 terminal nodes; to differentiate the joints, it used the eminence radius and the absolute superior joint space to anterior joint space ratio subordinate to the postglenoid process height. The tree model accounted for 31.4% of the likelihood (Rescaled Cox and Snell R(2)) with 73.5% accuracy (sensitivity 82.6% and specificity 65.4%). Disk displacement without reduction joints had either deeper posterior fossa walls or posterior walls of average length combined with a superior-to-anterior joint space ratio of less than 0.83; this suggests a more open-wedge-shaped anterior joint space combined with a less-rounded articular eminence. In contrast, most disk displacement with reduction joints had shorter posterior fossa wall height combined with more equal or larger superior-to-anterior joint spaces. The logistic regression model was less accurate than the classification tree model (sensitivity 60.9%, specificity 66.7%) and accounted for only 9.9% of the likelihood (Rescaled Cox and Snell R(2)) and 63.6% accuracy. The postglenoid process height was the strongest differentiating factor in all models.

Hard tissue relationships revealed by central tomogram sections were able to model notable differences between disk displacement with and without reduction joints when examined as contingency-based multifactorial systems.