Five out of twenty four (5/24) failures had facet fractures on CT scan as in Figure 1. One of these was after a subsequent laminectomy and was considered a post laminectomy facet fracture (19). Three out of twenty four (3/24) had bilateral pars fractures with one of these being present at 2 levels pre-operatively. All twenty four patients had either distraction of the facets (figure 2) or conversely compression of the facets (figure 3) due to posterior concentration of stresses. Over-distraction varied from 50% to 100% of normal disc space. Fifteen out of twenty four (15/25) patients had leg pain in addition to back pain yet had no obvious areas of impingement on CT scan. Three of these patients were sent for diagnostic facet injections and all had some partial relief of the leg pain, but only for the duration of the local anesthetic. Six out of twenty four (6/24) had endplate fractures resulting in subsidence with all but one being less than 4mm of subsidence. .In patient 11, pictured in Figure 4, the subsidence fracture was approximately 30% of the vertebral body height causing posterior extrusion of the polyethylene component.

Figure 2

Figure 1: Facet fracture after Prodisc replacement in patient #15.

Figure 3

Figure 2: Facet distraction after Charite replacement in patient #20. Normal facets at level above are shown for comparison.

Figure 4

Figure 3: Facet compression after Charite replacement in patient #5: Normal facets at level above are shown for comparison.

Figure 5

Figure 4: Endplate fracture/subsidence in patient #11

A common clinical feature of all twenty four patients was severe back pain. Nineteen out of twenty four (19/24) patients also had leg pain with ten out of twenty four (10/24) being bilateral and nine out of twenty four (9/24) being unilateral. Five out of twenty four (5/24) patients have been surgically revised. Of these, three operative revision patients at this institution were contemplating suicide pre-operatively due to the pain.

In all patients presenting to UCI, a minimum of 6 weeks of physical therapy along with NSAIDS had been already tried prior to presentation at our institution. As noted in Table 1, some patients already had epidural steroid injections (ESI), facet blocks, and in a few cases, facet rhizotomies with only patient #5 noting temporary relief. Four (4) patients underwent diagnostic facet blocks under our care which were successful in temporarily relieving pain. Three (3) of these underwent revision at our institution, and one is currently scheduled for this. All patients remained on daily narcotics for pain control.

Our rationale for the surgical treatment was based on three concepts. Firstly, it was postulated that there was a posterior transfer of load that caused compression of the facets and subsequent severe local back pain. Even when the replacements seemed in good position, the lack of any shock absorption as in normal disc motion was thought to add to this posterior transfer of load. Secondly, in cases where the facets were distracted and the patients had leg pain, the capsular stretch along with traction on the nerve root and dorsal ramus was suspected to be the source of pain. Clinical evidence that support these concepts was the relief of back and leg pain temporarily with a diagnostic facet injection in some patients as indicated in Table 1. Thirdly, the transfer of load posteriorly was thought to be the reason for the “subsidence” of the superior part of the disc replacement into the superior vertebral body. This was considered, as discussed below, to realistically be a fracture of a portion of the endplate. As such, it was thought to be a source of pain, as any fracture would. The observation that it almost always occurred in the posterior third of the vertebral body is consistent with the idea of posterior transfer of load.

The surgical treatment would, therefore, be to address the above concepts. It was felt that the compressed facets, which were often fractured and not healed - namely a non-union - needed to be excised or at least unloaded. The distracted facets required the release of the tension on the capsule, nerve root, and dorsal ramus by either excision of the facet or at least some reduction of the distraction by compression. It was reasoned that the endplate fractures needed to be unloaded to allow healing, or least relief of pain. These actions would be accomplished during an instrumented posterior fusion. If the premises for the sources of pain were correct, and these were addressed in the fashion described, then removal of the prosthesis would not be necessary. Figure 5 is an example of a patient with a facet fracture after disc replacement. Treatment, shown in Figure 6, was with excision of the facet followed by a posterior fusion. This resulted in excellent relief of the low back and thigh pain. Bone morphogenetic protein in a compression resistant matrix was used as the graft material in the lateral gutters for the postero–lateral fusion. Instrumentation in all of the cases was in the form of pedicle screw fixation. Figure 7 shows an example of distracted facets. The normal facets above are pictured in Figure 8 for comparison. This was patient #23 who had bilateral leg pain that was relieved after facet excision, instrumentation with slight compression, and then fusion. Although one would think this is only minimally appearing distraction, it was nevertheless unstable at the time of surgery. Grasping the posterior spinous process at this level revealed surprising gross instability of the sort one would expect in an unstable spondylolisthesis. Yet, it was only the distraction of the facets joints to this degree that was abnormal. However, it should be remembered that this was imaged with the patient in the supine position.

Figure 6

Figure 5: Facet fracture at L5-S1 in patient #20.

Figure 7

Figure 6: Fusion of patient #20.

Figure 8

Figure 7: Distracted facets in patient #23.

Figure 9

Figure 8: Normal facets of patient #23.

After seeing these patients and trying to piece together the causes of pain, a number of parameters became apparent as primary etiologic factors based on observation and the literature.

The center of rotation (COR) of the normal lumbar segment is the center of a circle formed by the facets as seen in Figure 9. ₁₁ As the disc compresses under a load, the vertebral body goes down as seen in Figures 10 and 11. Since all motions in the spine are well known to be coupled, the flexion and extension is accompanied by rotation as Figures 10 and 11 accurately depict. The COR for this combined motion is the COR of the facets in Figure 9. Because there are no collateral ligaments, there is no restriction to this side-to-side rotational motion. (The anterior and posterior longitudinal ligaments restrict motion in flexion and extension only, not side-to-side.) This combined motion is viewed from above in Figures 12 and 13. Essentially, Figures 10-13 show the motion dictated by the facet anatomy. Although coupled motion is beyond the scope of this article, one key concept should be mentioned. Coupling the motion decreases the amount of compression shown, but does not change the side to side motion of the vertebral bodies dictated by the facet anatomy.

Figure 10

Figure 9: Center of rotation (COR) of a lumbar segment.

Figure 11

Figure 10: Segment in neutral.

Figure 12

Figure 11: Segment in coupled flexion and rotation.

Figure 13

Figure 12: Segment in neutral (top view).

Figure 14

Figure 13: Segment in coupled flexion and Rotation.

The presence of anterior and posterior longitudinal ligaments, and the absence of collateral ligaments, facilitates the side to side motion shown above. The authors considered the anterior longitudinal ligament as a checkrein to excessive extension in the disc space. In performing a disc replacement, this ligament is incised. It is felt that this allows for a significant increase in extension beyond normal. Extrusion of the disc replacement as a result of this and the ligaments absence as containment may subsequently occur as in Figure 14. It is also thought that abnormally high loads occur posteriorly while concurrently being decreased anteriorly adding to this occurrence. The polyethylene core may be expelled either with or without the metal endplates. Catastrophic results may ensue due to vascular injury.

Figure 15

Figure 14: Extrusion.

Motion preservation appears to be accomplished largely by removing the anterior longitudinal ligament. The resultant increase in extension is discussed frequently by advocates of disc replacement, but there is a notable paucity of discussion about the lack of increase in flexion and rotation. These latter motions are a greater part of daily activities than isolated extension.

The disc replacement design, in contrast to the normal human disc, assumes that all centers of rotation are anterior to the spinal canal in the disc space. With the replacement in place, there are now two centers that sum to disallow motion at all unless something “gives.” What “gives” depends on which center of rotation ‘wins” out over the other. If the disc replacement's center of rotation that is anterior dominates, then the facets which are posterior are damaged by abnormal movements. If the normal posterior center about the facets dominates, then the disc replacement, which is anterior, moves abnormally and subluxes or dislocates. In Figure 15, the false center of rotation artificially placed anteriorly by the disc replacement overcomes the normal posterior center. Subsequently, the facets undergo abnormal motion and loading which can lead to a fracture or rapid degeneration as in Figures 16 and 17. Here, the degree of facet degeneration that one would expect to occur over a period of years instead occurred over a period of 8 months after disc replacement. Local anaesthetic injection relieved all back and thigh pain temporarily in this patient. Subsequently, she underwent surgical treatment with facet excision and fusion that resulted in excellent pain relief.

Figure 17

Figure 16: Normal facets one month prior to disc replacement in patient #22.

Figure 18

Figure 17: Same facets as in figure 15 but 8 months after disc replacement. Note joint space loss and subchondral cysts.

{image:18}

In the patients in this study, in cases of severe distraction of the facets, there was often accompanying leg pain that lacked any radiographic evidence of impingement. This suggested that traction on the underlying nerve root may be a reasonable etiology. This could be either directly or via the dorsal ramus. The cases of pain relief after facet block in this situation appear to support this proposed etiology of the pain. As a result of this, the rationale for excision or compression of the facets surgically was formed. As discussed above, this was done to relieve the tension on the capsule, nerve root, and dorsal ramus. Leg pain relief was excellent when this was done (Table 1).

The complication described as “subsidence” of the metal endplate of the disc replacement into the vertebral body is more accurately described as a fracture of the vertebral endplate, in the authors' opinion as discussed above in the rationale for surgical revision. After all, one cannot imagine how the metal plate migrates into the body of the vertebrae unless the structural integrity of the bone is compromised - namely fractured. The artificial disc does not have any of the natural shock absorption that the healthy human disc does, and this is thought by the authors' to contribute to a posterior transfer of load.₁₄This affect is engendered by the excision of the anterior longitudinal ligament which leads to an increase in extension. Yet, the posterior longitudinal ligament remains intact. Thus, the vertebral body has the anterior loads decrease and the posterior loads increase. Add to this the increased forces of distraction of the disc space by a large disc replacement, and the groundwork for a subsidence fracture in this area is laid. All the cases of endplate fractures in this study occurred in the area of the posterior third of the vertebral body.

All previous biomechanical models of spinal motion used mathematical assumptions that were well known to be unreliable and erroneous for the small degrees of angular measurement of segmental spinal motion. As a result, all current disc replacements that are based on these models may be inherently flawed in their intended function. Instantaneous center of rotation is defined as a point on the plane having zero velocity at the instant considered.₁₈ Since Reuleaux₁₉ first published a graphical solution to find the center of rotation, as the point of intersection of perpendicular bisectors of the two lines joining the initial and final positions of two points on the plane. Instantaneous center of rotation and center of rotation have been used interchangeably based on the assumptions that the difference in the actual location of the center of rotation and instantaneous center of rotation of a joint are small and the differences decrease with the number of intervals into which the total range of motion is divided.₂₀ Due to its simplicity Reuleaux method is still one of the most commonly used techniques to locate the center of rotation. However, several studies have demonstrated not only its inaccuracy to locate the center of rotation for a small amount of angular displacement but also its high dependency on the marker conditions, which might not be able to optimize during the experimental testing.₂₀,₂₁ Several guidelines were also suggested to maximize its accuracy to locate the center of rotation.₂₂,₂₃ Spiegelman and Woo₂₁ developed a rigid-body method for finding centers of rotation and angular displacements of planar joint motion which has fewer constraints on locating the markers, resulting in more accurate determination than with the Reauleaux method. Among all of the currently-existing methods, however, there is a strong agreement that there is no technique to accurately locate the center of rotation for a small amount of angular displacement, such as in thoracolumbar spinal joints.