Hip dislocation is a major short-term complication after total hip arthroplasty (THA). One factor thought to reduce the risk for dislocation is head size. We constructed subject-specific computer models to study the effect of head size on risk for postoperative dislocation. Femoral and acetabular geometry was constructed after segmenting CT scans of nine hips. CAD models of THA components with four head diameters (28, 32, 36, and 44 mm) were virtually implanted. Hip capsular ligaments were simulated using rigid-body ellipsoids connected by non-linear springs. Posterior dislocation was simulated during a rise from a low chair; anterior dislocation was simulated during a pivot activity. Intraoperative stability tests were simulated for anterior or posterior dislocation. While rising from a low chair (posterior dislocation) and during the pivot activity (anterior dislocation), increasing head size significantly increased hip flexion angle at dislocation and generated higher dislocation moments. Larger heads reduced the risk for dislocation. Intraoperative stability tests detected the relative increased resistance to dislocation despite differences in the absolute magnitude of moments. This model can be useful preclinical tool for assessing design changes, the effect of component placement, and the activity-based risk for dislocation. © 2014 Orthopaedic Research Society. Published by Wiley Periodicals, Inc. J Orthop Res 32:1525–1531, 2014.
Hip dislocation is one of the most significant early complications after total hip arthroplasty (THA).[1-6] Although its etiology is multifactorial, prosthetic impingement is an important factor that is thought to reduce the risk.[3, 7-10] Advances in hard-on-hard bearings and crosslinked polyethylene liners have renewed interest in larger head sizes. Increasing head–neck ratio theoretically increases hip range of motion before impingement and therefore can reduce the potential for hip dislocation. We previously reported that increasing head-size beyond 32 mm head diameter did little to improve range of motion because bony impingement became the limiting factor, especially when the components were at or near optimal orientation.
In vitro and computational models of hip dislocation have been used to study the effect of head size.[3, 11-15] Factors affecting dislocation include angle at impingement, angle at dislocation, and the moment resisting dislocation. The angle at impingement (or range of motion) is often used as a dislocation risk factor.[9, 12, 13, 16-18] Models using the angle at prosthetic impingement as a marker for risk of dislocation reported increased hip range with larger head sizes primarily because of the increase in head–neck ratio.[11, 15] Conversely, when bony impingement is also simulated, we and others found that head size >32 mm did little to improve range of motion before impingement, especially when components were placed within the optimum range.[3, 13, 19]
In contrast to the angle at impingement, the angle at dislocation does increase with head size even when bone-on-bone impingement restricts range of motion. This result is due to the increased distance the larger head has to travel to dislocate: the so called “jump” distance. Dislocation studies in anatomic replicas of the femur and pelvis and in cadavers also found an increase in the maximum flexion before dislocation with increasing head size.[7, 13]
Hip dislocation is also dependent on capsuloligamentous stability. Apart from increasing head–neck ratio and potentially increasing range of motion before impingement, larger heads also increase the distance necessary for the head to translate before dislocating out of the socket and may therefore have an additional benefit even if bony impingement were limiting range of motion.
To test this hypothesis, we constructed a computer model of dynamic hip dislocation using rigid body mechanics and contact to study the effect of head size on the risk for postoperative dislocation. Most surgeons perform an intraoperative stability test to determine potential for dislocation. Our secondary objective was to test the hypothesis that an intraoperative stability test would predict the potential for postoperative dislocation during a high-risk activity such as rising from a low chair.
With Institutional Review Board approval, we segmented CT scans of nine hips using MIMICS (Materialise, Leuven, Belgium) to construct the femoral and acetabular geometries. CAD models of THA components (Secur-Fit HA) were provided by Stryker Orthopaedics (Mahwah, NJ) in four head diameters (28, 32, 36, and 44 mm) and were combined with a femoral component with a neck diameter of 12.5 mm and neck-stem angle of 132°. These components were aligned to the geometry of each hip using a previously described technique.[3, 19] We chose an orthogonal coordinate system based on the recommendation of the International Society of Biomechanics and oriented the pelvis so that the anteroposterior iliac spines were level and in the same frontal plane as the pubic symphysis (no lordosis or pelvic obliquity). We located the center of the natural femoral head by fitting a sphere to the articular surface of the head and used the same method to locate the center of the natural acetabulum. Then, we aligned the head center with center of the acetabulum, orienting the long axis of the femur (the line joining the femoral head center to the midpoint of the intercondylar notch) perpendicular to the transverse axis of the pelvis (defining neutral hip abduction) and parallel to the frontal plane of the pelvis (defining neutral hip flexion). We used a line passing through the medial and lateral epicondyles to define neutral femoral rotation. The center of the hip axis was located at the head center. In neutral position, the x-axis was pointed in the anterior direction, the y-axis superiorly, and the z-axis toward the medial aspect of the patient's right side. We described flexion–extension around the pelvic z-axes (fixed to the pelvis) and axial rotation of the hip around the femoral y-axis (fixed to the femur). The axis for abduction–adduction was the floating axis (perpendicular to both the y- and z-axes). We defined abduction of the acetabular cup from the horizontal line around the x-axis of the pelvis, whereas the true anteversion was the rotation of the cup around y-axis of the pelvis (as opposed to apparent radiographic anteversion, which was rotation of the cup about its abducted axis).
We virtually implanted these models into each of the reconstructed anatomic models under the direction of a joint arthroplasty surgeon as previously described. In the model, we resected the bony anatomy of the head and neck 13.5 mm above the lesser trochanter at 45° to the vertical axis. We reamed the natural acetabulum by creating a sphere (sphere diam = outer cup diam). This sphere was subtracted (using Boolean subtraction) from the natural acetabulum to generate a surface resembling a reamed acetabulum. We virtually implanted the hip components in the pelvis and the femur. For initial placement, we chose the implant sizes and the locations of the cup relative to the pelvis and the stem relative to the femur. We used the equivalent of preoperative radiographic templating to select a prosthetic size for best fit of the femoral stem in the canal and the acetabular component in the acetabulum and for initial implant position. We maintained the original hip center of rotation (center of the sphere defining the native acetabulum) and limb length (vertical level of tip of the lesser trochanter relative to the hip center), while aligning the long axis of the stem with the long axis of the intramedullary canal of the femur. The acetabular center was unchanged, since the center of the acetabular component was aligned with the center of the natural acetabulum during the simulated implantation. The cup was abducted 45° and anteverted 20° using the previously defined pelvic coordinate system. The outer diameter of the acetabular liner was adjusted to fit the acetabulum for each subject, while the inner diameter matched to each corresponding head size.
We used a rigid body dynamics solver (MSC.ADAMS, MSC Software, Santa Ana, CA) to simulate hip dislocation. Contact between the femoral head and the acetabular liner was simulated with a coefficient of friction of 0.09. We have validated the accuracy of the contact detection algorithm in predicting range of motion before impingement in cadaver tests.
Ligament contact wrapping was implemented to simulate the superior and inferior iliofemoral, and the ischiofemoral ligaments to represent the hip capsule (Fig. 1). The iliofemoral ligament originated from the anterior inferior iliac spine and acetabular rim and divided into superior and inferior sections that inserted proximally and distally along the intertrochanteric line, respectively, on the femur. The origin of the ischiofemoral ligament was at the ischial rim of the acetabulum and the insertion at the posterior aspect of the femoral neck followed a spiral pattern distally around the joint. We used rigid-body ellipsoids connected by non-linear springs to construct these ligaments. The non-linear springs were modeled as quadratic for strain ≤6%; higher strains were modeled as linear[25, 26]:
where f, force; k, ligament stiffness; and e, ligament strain.
The stiffnesses used for the linear phase was 3,560, 1,706, and 719 N/mm for the superior iliofemoral ligament, inferior iliofemoral ligament, and ischiofemoral ligament, respectively.
The ellipsoids were used to detect contact between the ligaments and the hip components to simulate capsular wrapping and to compute resistance to dislocation before failure. Ligament deformation during wrapping can alter the local cross-sectional area, but requires finite element analysis for accurate simulation.[20, 27] We used contact between rigid-body ellipsoids and the femoral head as a simplified model of ligament wrapping. This form of wrapping is often used in rigid body musculoskeletal models. This study was focused on the net forces acting on the femoral head resisting dislocation, rather than the effect on the ligaments. We therefore traded lower fidelity in ligament modeling for computational speed, which made it possible to efficiently run models representing different patients across many conditions.
The activities simulated are listed in Table 1. Posterior dislocation was tested under three conditions. To replicate an intraoperative test, we first placed the hip in 100° of flexion and 15° of internal rotation. The hip was then adducted until the head dislocated. We repeated the test after removing the capsular ligaments to assess their contribution to the peak-resisting moment and because intraoperative testing is typically conducted before repairing the capsule. To simulate a rise from a low-seated chair, an activity at high risk for posterior dislocation, the hip was initially placed in 90° of flexion, 10° of internal rotation, and 10° of adduction. The hip was then progressively flexed (to simulate trunk flexion) until the head dislocated.
|Direction of Dislocation||Activity Type||Capsular Ligaments|
We also tested anterior dislocation under three conditions. To replicate an intraoperative test, we first placed the hip in 15° of extension and then rotated it externally until the head dislocated. The test was repeated after removing the capsule to simulate the intraoperative condition. Weight-bearing on the affected limb followed by rotating the torso away from the affected hip has a high association with anterior dislocation. To simulate the pivot activity, the hip was initially placed in 15° of extension and 5° of abduction, then progressively rotated externally (to simulate trunk rotation) until the head dislocated anteriorly.
For each activity (Table 1), peak-resisting moment, angle of impingement, angle of dislocation, and individual ligament forces were measured for each head size.
We used repeated measures ANOVA (Systat®, Systat Software, Chicago, IL) to test for significant differences in angle at impingement, angle at dislocation, and the peak moment resisting dislocation among the three head sizes. Bonferroni-corrected pair-wise comparisons were used to determine differences among head sizes.
Effect of Capsular Ligaments
The linear stiffness for each ligament was obtained from Hewitt et al., in which the mechanical testing of individual hip ligaments was reported. For validation of our implementation of these ligament properties, we compared the results of whole joint distraction simulation with those reported experimentally. The predicted force-displacement during distraction closely matched the average of the experimentally measured force-displacement (Fig. 2). The presence of capsular ligaments significantly increased the peak moment resisting dislocation, averaging a 2-fold increase for anterior dislocation and a 2.4-fold increase for posterior dislocation (p < 0.001, Fig. 3).
Intraoperative Stability Tests
During the intraoperative stability test for posterior dislocation, the hip adduction angle at impingement increased significantly when head diameter was increased from 28 to 32 mm, but not with the further increase to 44 mm (Fig. 4). Conversely, the hip adduction angle at dislocation was significantly higher in each of the larger heads (p < 0.05, Fig. 4). The larger heads also generated a higher moment resisting dislocation (p < 0.002, Fig. 5) during the intraoperative stability test for posterior dislocation. During the intraoperative stability test for anterior dislocation, the hip external rotation angle at impingement did not increase significantly with head size. However, the hip external rotation angle at dislocation was significantly higher in larger heads (p < 0.02, Fig. 4), although these differences were small. Larger heads also generated a higher moment resisting dislocation (p < 0.005, Fig. 5).
While rising from a low chair, increasing the head size significantly increased hip flexion angle at dislocation (p < 0.05, Fig. 4) and also generated a higher peak moment resisting dislocation (p < 0.05, Fig. 5). For anterior dislocation during the pivot activity the larger head sizes also increased the external rotation range of motion before dislocation (p < 0.05, Fig. 4). However, only the 36 mm head generated a significantly greater peak moment resisting dislocation (p < 0.05, Fig. 5). Head size had no effect on impingement angle during the chair rise or the pivot activity.
We constructed a computer model of hip dislocation based on rigid body mechanics and contact, incorporating the effect of local bony anatomy and capsular restraints on dynamic posterior dislocation. Our primary hypothesis was that head size would increase the resistance to dislocation. The potential for dislocation is assessed by most surgeons by performing an intraoperative stability test. Our secondary hypotheses were that the resistance to posterior dislocation during an intraoperative stability test would be consistent with that computed for rising from a low chair, an activity that is at high risk for posterior dislocation, and that resistance to anterior dislocation during an intraoperative stability test would be consistent with the standing pivot test, an activity that is at high risk for anterior dislocation.
Several in vitro and computational models of hip dislocation have been used to study the effect of head size.[3, 11-14] Common predictors of hip dislocation include angle at impingement, angle at dislocation, and the moment resisting dislocation. The angle at impingement is often used as a risk factor.[9, 12, 13, 16-18] Models using angle at impingement as a marker for dislocation risk reported increased hip range of motion with larger head sizes primarily because of the increased head–neck ratio.[11, 15] Conversely, increasing head size beyond 32 mm does little to improve range of motion before impingement, especially when the components are placed within the optimum range.[3, 13] In contrast to the angle at impingement, the angle at dislocation did increase with head size even when bone-on-bone impingement restricted range of motion. This outcome is due to the increased distance the larger head has to travel to dislocate. Studies of hip dislocation in anatomic replicas of the femur and pelvis as well as in cadavers also found increased maximum flexion before dislocation with increasing head size.[7, 13] Finite element models have also reported increase in peak dislocation resisting moment with larger head sizes, again attributable to increased head–neck ratio. In general, the posterior dislocation tended towards a greater proportion of hips undergoing component impingement, while anterior dislocation tended towards a greater proportion of hips undergoing bony impingement (Table 2).
|Head Size (mm)||Rising From a Low Chair||Intraoperative Posterior Test||Pivot Test||Intraoperative Anterior Test|
We validated the accuracy of the contact detection algorithm in predicting range of motion before impingement in cadaver tests. We compared the biomechanical performance of our simulated hip ligaments against experimentally measured biomechanics of the whole hip joint during distraction simulation. The close agreement between predicted force-displacement during distraction and the average of the experimentally measured force-displacement served to validate our selection of ligament properties. Our computer-generated peak torque at dislocation (with and without hip ligaments) was within the range reported during cadaver testing of hip dislocation after THA. We also compared the results of our rigid-body dynamics approach to a study simulating the behavior of the hip capsule during dislocation using explicit finite element analysis. Simulating contact between the femoral head and the capsule with finite element analysis increased the peak moment before dislocation by more than threefold. In our rigid-body dynamics model, we also found an up to 2.5-fold increase in peak moment resisting dislocation. These significant differences in dislocation moment emphasize the importance of simulating the capsuloligamentous structures.
Head size had a significant effect on maximum flexion angle before dislocation and on the peak dislocation moment while rising from a low chair. Greater flexion and higher peak dislocation moment indicate increased hip stability. Thus, larger heads appear to significantly reduce the potential risk for posterior dislocation in this high-risk activity. These results are supported by our clinical report of reduced dislocation rates in THAs with 32 and 36 mm heads compared to 28 mm heads.
A common test for posterior dislocation is measurement of the maximum range of adduction possible with the hip in 100° flexion and 15° of internal rotation. For anterior dislocation, surgeons often extend the hip to 15° and then assess the stability in external rotation. However, the validity of these tests and their sensitivity in identifying risk for postoperative dislocation are unknown. Our results indicate that the absolute magnitudes of peak dislocation moment during intraoperative stability testing were somewhat different from the postoperative chair rise, especially in the absence of capsular ligaments. When ligaments were included during intraoperative anterior stability testing, the dislocation moments were similar to the postoperative pivot test. This was likely due to the similarities in hip position between the intra- and postoperative conditions. When testing for stability against posterior dislocation, however, hip flexion is often limited by patient position and surgical support equipment. Therefore, stability during hip adduction is used as a surrogate indicator. Including ligaments during intraoperative posterior stability testing did increase the dislocation moment, although not to the same extent as for anterior stability testing. Nevertheless, the relative differences between the head sizes were similar even in the absence of capsular ligaments. This finding reinforces the validity and clinical relevance of a passive intraoperative test in assessing the relative contribution of head size in increasing the range of motion before impingement or dislocation.
A weakness of our study was that while we incorporated subject-specific bony anatomy and hip ligament insertions, the same material properties were used for all subjects. These material properties were those reported for normal capsuloligamentous structures. Ligament deformation during wrapping can alter the local cross-sectional area, but requires finite element analysis for accurate simulation.[20, 27] We used contact between rigid-body ellipsoids and the femoral head as a simplified model of ligament wrapping. This form of simplified wrapping is often used in rigid body musculoskeletal models. We focused on the net forces acting on the femoral head resisting dislocation, rather than the effect on the ligaments. We therefore traded lower fidelity in ligament modeling for computational speed, which made it possible to efficiently run models representing different patients across many conditions.
In summary, we developed subject-specific computer models of hip dislocation that incorporated bony impingement and the effect of the capsule in stabilizing the hip and controlling dislocation. Larger heads generated significantly higher peak moments resisting dislocation and increasing the range of motion before dislocation. The intraoperative tests for stability, typically used during THA surgery, also detected the increased resistance to dislocation despite the difference in the absolute magnitude of forces and moments. This model serves as a useful preclinical tool for assessing design changes, the effect of component placement, the activity-based risk for dislocation, and for gaining insights into mechanisms contributing to hip dislocation.
Two of the authors received support from Stryker Orthopaedics for research projects not related to this study.