Current finite element modeling techniques utilize geometrically inaccurate cartilage distribution representations in the lumbar spine. We hypothesize that this shortcoming severely limits the predictive fidelity of these simulations. Specifically, it is unclear how these anatomically inaccurate cartilage representations alter range of motion and facet contact predictions. In the current study, cadaveric vertebrae were serially sectioned, and images were taken of each slice in order to identify the osteochondral interface and the articulating surface. A series of custom-written algorithms were utilized in order to quantify each facet joint's three-dimensional cartilage distribution using a previously developed methodology. These vertebrae-dependent thickness cartilage distributions were implemented on an L1 through L5 lumbar spine finite element model. Moments were applied in three principal planes of motion, and range of motion and facet contact predictions from the variable thickness and constant thickness distribution models were determined. Initial facet gap thickness dimensions were also parameterized. The data indicate that the mean and maximum cartilage thickness increased inferiorly from L1 to L5, with an overall mean thickness value of 0.57 mm. Cartilage distribution and initial facet joint gap thickness had little influence on the lumbar range of motion in any direction, whereas the mean contact pressure, total contact force, and total contact area predictions were altered considerably. The data indicate that range of motion predictions alone are insufficient to establish model validation intended to predict mechanical contact parameters. These data also emphasize the need for the careful consideration of the initial facet joint gap thickness with respect to the spinal condition being studied.

Utilization of a more physiologic cartilage thickness distribution in FE models will result in improved representation of cervical spine kinematics and increased predictive power. The consistency observed in the thickness distribution function in this study indicates that such a representation can be generated relatively easily.