Joint Contact Stress Improves in Dysplastic Hips After Periacetabular Osteotomy but Remains Higher than in Normal Hips

Abstract

Aim:

The purpose of this study was to use computational modeling to determine if surgical correction of hip dysplasia restores hip contact mechanics to those of asymptomatic, radiographically normal hips.

Methods:

Discrete element analysis (DEA) was used to compute joint contact stresses during the stance phase of normal walking gait for 10 individuals with radiographically normal, asymptomatic hips and 10 age- and weight-matched patients with acetabular dysplasia who underwent periacetabular osteotomy (PAO).

Results:

Mean and peak contact stresses were higher (p < 0.001 and p = 0.036, respectively) in the dysplastic hips than in the matched normal hips. PAO normalized standard radiographic measurements and medialized the location of computed contact stress within the joint.

Mean contact stress computed in dysplastic hips throughout the stance phase of gait (median: 5.5 MPa, [IQR: 3.9-6.1 MPa]) did not significantly decrease after PAO (3.7 MPa, [IQR: 3.2-4.8]; p = 0.109), and remained significantly (p < 0.001) elevated compared to radiographically normal hips (2.4 MPa, [IQR 2.2-2.8 MPa]). Peak contact stress demonstrated a similar trend. Joint contact area during the stance phase of gait in the dysplastic hips increased significantly (p = 0.036) after PAO from 395 mm² (IQR: 378-496 mm²) to 595 mm² (IQR: 474-660 mm²), but remained significantly smaller (p = 0.001) than that for radiographically normal hips (median 1120 mm², IQR: 853-1444 mm²).

Conclusion:

While contact mechanics in dysplastic hips more closely resembled those of normal hips after PAO, the elevated contact stresses and smaller contact areas remaining after PAO indicate ongoing mechanical abnormalities should be expected even after radiographically successful surgical correction.

INTRODUCTION

Hip dysplasia is a complex deformity of the acetabulum and the femur that results in joint instability and elevated intra-articular mechanical stress.^{1, 2} These abnormal mechanics cause significant hip pain and disability in young, active patients, and can lead to development of osteoarthritis of the hip in young adulthood.³ Periacetabular osteotomy (PAO) is a well-described procedure to correct acetabular deformity in adult hip dysplasia. The goal of the procedure is to reorient the articular surface of the acetabulum to improve deficits in coverage of the femoral head, increase and medialize joint contact area, and reduce joint contact stress.⁴ In theory, this provides a more favorable biomechanical environment for the entire hip joint, thereby relieving pain and reducing the risk of developing osteoarthritis. However, the 2D radiographic measures typically used to evaluate severity of dysplasia and to assess the quality of correction (lateral center edge angle [LCEA], Tönnis angle, extrusion index, etc.) do not provide information about how PAO alters functional hip mechanics.

Discrete element analysis (DEA) is a computational modeling technique that utilizes rapidly executing, highly numerically stable computations to calculate joint contact stress from relatively simple surface models. Joint contact stresses calculated using DEA have been shown to relate to future development of osteoarthritis,^5-7 making this technique useful for evaluating joint mechanics as related to future development of disease. Multiple studies over the past several decades have used DEA to evaluate joint contact mechanics in dysplastic hips,^8-10 confirming that joint contact stress is higher and contact areas are reduced in dysplastic hips relative to normal hips.^{1, 2} Those findings have led to continued efforts to identify patient-specific, biomechanically-based, optimal surgical corrections to reduce contact stress^11-15 and relate postoperative changes in contact mechanics to improvements in patient outcomes.^{16, 17} While this body of computational work has focused on identifying and optimally improving abnormal contact stresses, there is extremely limited information about how the contact stresses that can be achieved after PAO in a dysplastic hip joint compare to contact stresses in a radiographically “normal” hip joint. The purpose of this study was to use DEA to calculate contact mechanics in patients with acetabular dysplasia before and after PAO, and compare those stress patterns to those in age/weight-matched patients with radiographically normal and asymptomatic hip joints. We hypothesized that PAO would reduce contact stress and increase contact area in the dysplastic hips, but that contact mechanics after surgical treatment would differ significantly from those of radiographically normal hip joints.

METHODS

After institutional review board approval, a series of 10 individuals who underwent a pelvis CT scan as part of a routine trauma workup were retrospectively identified through chart review based on meeting the following criteria: 1) good quality supine AP pelvis radiograph without evidence of acetabular dysplasia defined by standard radiographic measurements;¹⁸ 2) alpha angle <55 degrees on standing AP pelvis radiograph; 3) 18-45 years old at the time of presentation; 4) no report of hip pain or hip or pelvis injury. Consecutive record review was performed until 10 cases were identified. Radiographic measurements were performed by a board-certified orthopaedic surgeon (MW).

10 age and weight-matched individuals were selected from a historical database of 112 patients (135 hips) with acetabular dysplasia that were treated with PAO¹⁹ by a single surgeon at our institution between January 2003 and April 2010. All patients had preoperative and immediate postoperative CT scans available. Indications for PAO in this cohort were LCEA <25 degrees with hip pain that was not improved by conservative treatments, including physical therapy, activity modification, intra-articular corticosteroid injections, or anti-inflammatory medications. Patients in this database with a history of Perthes disease, slipped capital femoral epiphysis, or whom had acetabular retroversion were excluded as potential matches.

Femoral, pelvic, and sacral bony geometry was segmented from each CT scan using a previously described, semi-automated workflow.^{20, 21} Cartilage was approximated by projecting the femoral and acetabular subchondral bone surfaces into the joint space a uniform distance equal to half of the mean joint space around the weightbearing lunate surface. At any location in which this method resulted in overlap of the projected surfaces, projections were adjusted to be half the distance between the bone surfaces at that location. To achieve a continuous, non-spherical, non-uniform thickness cartilage representation, which is critical for accurate contact stress calculations,²² the projected surfaces were smoothed using a custom algorithm that combines radial and nearest neighbor smoothing²³ with removal of any resultant overlap with the opposing cartilage surface. Cartilage surfaces were assigned isotropic linear-elastic material properties (E = 8 MPa, ν = 0.42).²⁴

Discrete element analysis (DEA)^{25, 26} was utilized to compute the intra-articular contact stress in each hip model during the stance phase of a walking gait cycle. The average hip joint reaction forces and hip rotation angles measured in a series of patients with instrumented total hips^{27, 28} were discretized into 13 evenly distributed static positions spanning the stance phase of walking gait. This set of average joint reaction forces was scaled by study participant bodyweight for application to that given subject’s DEA model. DEA was performed utilizing a custom algorithm implementing a Newton’s method solver previously developed in MATLAB.²⁹

Radiographic measurements, DEA-computed contact stress, and DEA-computed contact area were compared between the preoperative and postoperative conditions in the dysplastic hips and between postoperative dysplastic and normal hips using two-sided Wilcoxon rank-sum tests in SAS 9.4 (SAS Institute Inc, Cary, NC, USA). Wilcoxon matched-pairs signed rank tests were used to compare demographic data between patient groups. All data are presented as medians with interquartile range (IQR), and statistical significance was set at p ≤ 0.05 with a Holm-Bonferroni correction for multiple comparisons.

RESULTS

The median ([IQR]) difference between individuals with normal hip anatomy and the associated matched individuals with hip dysplasia was 4.5 years of age ([3.3-5 years difference]) and 4.1 kg ([2.4-8.1 kg difference]) of body mass (Table 1). There was no significant difference between groups in terms of age (p = 0.459), body mass (p = 0.557), acetabular cartilage thickness (p = 0.375), or femoral head cartilage thickness (p = 0.625).

Table 1.

Demographic and anatomic data for the individuals with normal hip anatomy meeting the inclusion criteria and the matched individuals with dysplastic hip anatomy. Cartilage thickness is the average cartilage thickness in the DEA models after the projection method using half of the joint space width and the subsequent smoothing.

	Age (years)		Weight (kg)		Acetabular Cartilage Thickness (mm)		Femoral Head Cartilage Thickness (mm)
Patient Pair	Normal	Dysplastic	Normal	Dysplastic	Normal	Dysplastic	Normal	Dysplastic
1	17	21	55.1	58.8	0.95	0.82	0.98	0.85
2	19	14	86.5	83.7	0.88	0.73	0.90	0.75
3	24	28	88.0	88.1	0.88	0.82	0.92	0.89
4	26	31	108.0	107.4	1.20	0.85	1.28	0.90
5	33	24	131.5	136.0	1.17	1.01	1.20	1.08
6	37	30	86.2	83.9	0.75	1.23	0.81	1.30
7	37	32	131.5	117.2	1.09	1.00	1.12	1.05
8	38	41	90.7	99.9	1.01	0.49	1.05	0.57
9	38	40	135.2	127.8	0.98	1.32	1.02	1.37
10	44	41	124.7	116.4	1.17	1.24	1.17	1.31
Median	35	30.5	99.3	103.6	0.99	0.92	1.04	0.98
IQR	[25, 38]	[25, 38]	[86.9, 129.8]	[84.9, 117.0]	[0.90, 1.15]	[0.82,1.17]	[0.94, 1.16]	[0.86, 1.24]

Median LCEA of the dysplastic group was 10 degrees ([7-17 degrees]), which increased significantly (p < 0.001) postoperatively to 31 degrees ([25-35 degrees]) (Table 2). This amounted to a per-patient median increase in lateral coverage of 19 degrees ([17-24 degrees]) after a PAO. Similarly significant improvements were seen in median Tönnis angle (21 degrees preoperative versus 3 degrees postoperatively, p < 0.001) and median extrusion index (0.35 preoperatively versus 0.16 postoperatively, p < 0.001) after PAO. There were no significant differences between dysplastic hips after PAO and normal hips in terms of median LCEA (31 degrees postoperatively versus 31 degrees in normal hips, p = 0.925), Tönnis angle (3 degrees postoperatively versus 2 degrees in normal hips, p = 0.752), or extrusion index (0.16 postoperatively versus 0.15 in normal hips, p = 0.697), indicating that PAO provided adequate radiographic correction of the previous dysplastic deformity (Table 2).

Table 2.

Radiographic measurements of the dysplastic deformity of the matched hip dysplasia patients preoperatively and after PAO. The same radiographic measures were made for the normal hips to ensure that postoperative correction was similar to normal anatomy.

	LCEA (degrees)			Tönnis Angle (degrees)			Extrusion Index
Patient Pair	Dysplastic	Post PAO	Normal	Dysplastic	Post PAO	Normal	Dysplastic	Post PAO	Normal
1	12	30	26	22	0	5	0.35	0.17	0.23
2	17	23	25	13	7	7	0.35	0.20	0.24
3	8	36	29	23	8	4	0.41	0.07	0.17
4	21	31	31	8	2	1	0.30	0.16	0.14
5	7	23	25	24	14	10	0.33	0.16	0.22
6	3	23	34	23	3	0	0.40	0.17	0.11
7	18	36	35	17	−4	0	0.27	0.14	0.12
8	15	40	30	12	−5	2	0.29	0.08	0.20
9	7	29	35	20	7	1	0.42	0.19	0.04
10	7	33	36	23	2	0	0.40	0.11	0.05
Median	10	31	31	21	3	2	0.35	0.16	0.15
IQR	[7, 16.5]	[24.5, 35.3]	[26.8, 34.8]	[14, 23]	[0.5, 7]	[0.3, 4.8]	[0.31,0.40]	[0.12, 0.17]	[0.11, 0.22]

PAO moved the center of the contact patch inferiorly (median: 3.21 mm, [IQR 1.77-5.64 mm]) and medially (2.33 mm, [1.78-2.96 mm]), but minimally in the anterior/posterior direction (0.33 mm posteriorly, [1.16 mm posteriorly - 0.96 mm anteriorly]) relative to the preoperative dysplastic condition. This movement medialized the spatial location of the contact stress patch on the acetabular cartilage surface (Figure 1) to a location that was not significantly different than in normal hips (p=0.695 for both inferior and posterior directions, and p=0.065 in for the medial direction). Mean contact stress (Figure 2) was significantly higher (p < 0.001) in the dysplastic hips preoperatively (5.5 MPa, [3.9-6.1 MPa]) than in normal hips (2.4 MPa, [2.2-2.8 MPa]). Following PAO, mean contact stress decreased to 3.7 MPa ([3.2-4.8 MPa]), but this was not a significant decrease (p = 0.109) from the preoperative state and was still significantly higher (p = 0.0003) than the mean contact stress in normal hips. Peak contact stress during the stance phase of gait was higher (p = 0.036 – not significant after Holm-Bonferroni correction) in the dysplastic hips preoperatively (19.1 MPa, [17.0-33.7 MPa]) than in normal hips (14.6 MPa, [13.8-16.8 MPa]). Peak contact stress in dysplasia patients decreased after a PAO (to 16.1 MPa, [13.1-25.2 MPa]), though this was not statistically significant (p = 0.353). The absolute peak contact stress at any point in the gait cycle remained elevated after PAO (p = 0.631) relative to normal anatomy. While the timing of the instant of peak contact stress varied among patients and reduced the global significance between groups, the dysplastic hips had notably higher peak and mean contact stresses throughout most or all of stance phase, respectively, compared to the hips with normal anatomy (Figure 2).

Figure 1.

Contact stress maps of the patients with normal hip anatomy (left) and associated maps from the matched patient with hip dysplasia before (center) and after (right) PAO. These maps are from step 4 (of 13), near heel-strike of walking gait. Anterior is to the left and superior towards the top.

Figure 2.

Mean (left) and peak (right) contact stress during the stance phase of a walking gait cycle. Solid markers indicate group medians and error bars indicate interquartile range at each modeled increment of the gait cycle. Statistical significance is indicated with the following symbols: *dysplastic preop vs. normal. ^#dysplastic postop vs. normal. ^†dysplastic preop vs. dysplastic postop.

The average contact areas calculated reflected a similar pattern to that found for contact stress. Preoperative contact area was a median of 395 mm² ([IQR: 378-496 mm²]) in the dysplastic hips, which was significantly smaller (p < 0.001) than the area calculated in the normal hips (1120 mm², [853-1444 mm²]; Figure 3). Contact area in the dysplastic hips did increase significantly (p = 0.036) to a median of 595 mm² ([474-660 mm²]) after a PAO, but this was still significantly (p = 0.001) smaller than for normal hips (Figure 3). Contact areas were also converted to percentages of the total acetabular area to account for differences in acetabulum size between dysplastic (2816 mm², [2668-3038 mm²]) and normal hips (3272 mm², [2986-3482 mm²]). Preoperatively, dysplastic hips had a median of 14.4% (12.4%-17.3%]) of the acetabulum in contact over the stance phase of gait, which was significantly (p < 0.001) less than that calculated for normal hips (34.1%, [30.8%-40.2%]). The percentage of the acetabulum in contact during stance phase of gait did increase to a median of 21.7% ([17.2%-24.2%]) after a PAO, but this was not significantly (p = 0.063) different from the preoperative state and was still significantly (p = 0.002) smaller than the portion of the acetabulum in contact for normal hips.

Figure 3.

Hip joint contact area over the stance phase of gait. Preoperative contact areas were significantly reduced in the dysplastic hips compared to the normal hips. PAO increased contact area relative to the preoperative state, but even after PAO the treated dysplastic hips had contact areas that were significantly smaller than the normal hips. Statistical significance is indicated with the following symbols: *dysplastic preop vs. normal. ^#dysplastic postop vs. normal. ^†dysplastic preop vs. dysplastic postop.

DISCUSSION

The goal of PAO is to restore stability and mitigate abnormal mechanical forces in a dysplastic hip, with a broader goal of reducing pain and preventing future joint degeneration. The findings of this computational modeling study suggest that despite clinically acceptable surgical correction of an existing dysplastic hip deformity, reorienting the acetabulum with a PAO does not restore a “normal” mechanical state. Patients with radiographically normal hips had significantly larger joint contact area than dysplastic hips both pre- and post-operatively, and while PAO did alter contact stress during gait, it did not reduce contact stress to levels found in normal hips.

We found dysplastic hips had peak contact stresses that were a median of 59% greater than the matched normal hips (IQR: 11% to 153%), which is similar to trends in previous work of dysplastic hips having on average 23-33% higher contact stresses than normal hips.¹¹ Similarly, in our dysplastic patient cohort, PAO reduced peak contact stress by a median of 12% (IQR: 5% to 26% reduction), which includes two patients with 4% and 9% increases in peak stress. This parallels findings of Armiger, et al., who reported that on average PAO reduced peak contact stress by a factor of 1.7, but also increased peak contact stress in one of the reported patients.¹⁷

Peak contact stress calculated in this work was generally higher than the 2-7 MPa that has been reported using DEA or similar methods for analysis of dysplastic hips during walking.^{8, 11, 12, 17} Our higher stress magnitudes are attributed to our effort to create a more realistic, non-uniform thickness cartilage layer, which has been shown to substantially affect computed joint mechanics.^{22, 30, 31} This approach differs substantially from the thicker cartilage and more spherical articulations that have been assumed in many previous studies.^{10, 11, 17} Interestingly, our results using patient-specific DEA compare extremely well with results from studies that utilized finite element (FE) analysis, a more sophisticated computational modeling technique that permits incorporation of multiple tissue types, advanced material properties, and provides continuum mechanics output information. Contact stress patterns calculated in our patient cohorts were extremely similar, both in terms of peak stress magnitude during walking gait and in discontinuous stress distribution, to those reported by Zou, et al. and by Henak, et al. for dysplastic and normal patients.^32-34 Those studies report average peak contact stress in the range of 12-15 MPa, which agrees well with our peak contact stress magnitudes near 19 MPa. The similarity of our findings with those from models incorporating a much broader selection of tissue anatomy and more sophisticated computational calculations verifies the contact stress values upon which the conclusions of this work are based. Further, our conclusions are similar to those of Zhao, et al. who used FE models of a normal hip morphed into a dysplastic configuration to determine if simulated PAO could restore the mechanics of the original normal hip. As was found in our DEA study of actual dysplasia patients, simulated PAOs in that FE model improved abnormal mechanics from a dysplastic condition, but did not fully return the stress state in the hip to normal.³⁵

Despite the similarity in findings verifying our methodology, there were several limitations to this work. First was the use of DEA to evaluate the mechanical differences between patient groups. DEA is a computationally efficient methodology; however, the results are limited to contact stress and contact area, with no information about internal tissue mechanics, shear behavior, or underlying bone mechanical behavior. While non-linear definitions of cartilage behavior can be included,^{12, 30} advanced material properties such as poroelasticity cannot be implemented to model cartilage, and the requirement of a compressive material between two rigid bodies prevents inclusion of an acetabular labrum or joint capsule in the models.

Another significant limitation to this work was the lack of patient-specific movement patterns to apply to the models. It is known that individuals with dysplasia have altered gait mechanics,³⁶ and the specific gait mechanics that are modeled significantly affect DEA-computed contact stress.³⁷ As we did not have any patient-specific gait data from either cohort, we could not account for patient-specific pelvic incidence, femoral version, muscle strength, sex-related gait variations, or gait modification after surgery, and we therefore chose to load our models with a single representative gait pattern. While the specific pattern we used was derived from patients older than a typical dysplasia cohort,²⁷ its use is well precedented in computational modeling studies of the hip,^{12, 13, 30, 34, 38} and this use of a single loading pattern allowed for differences in contact stress to be identified that were primarily a result of surgically induced changes in joint congruity. Unfortunately, with our single post-operative CT scan, we were unable to model further changes in contact stress associated with altered joint congruity over time after PAO. Future studies which include patient-specific gait modifications after PAO would benefit from additional post-operative 3D imaging to enable investigations of patient-specific joint remodeling on contact mechanics and joint health.

We selected 10 patients in each study group primarily due to limitations identifying a cohort of trauma patients within the age range typical of young adult hip dysplasia patients who did not have a hip injury/fracture and who also had the necessary quality CT scan available. This small group size encompassed a wide range of ages and dysplastic deformities. While our matching was effective in accounting for this, larger group sizes would have potentially allowed for stronger statistical comparisons, as well as determination of factors such as radiographic deformity or age that could have implications for the specific contact stress patterns present in particular subsets of patients. Dysplasia patients were preferentially matched to radiographically normal individuals based on bodyweight (rather than BMI), with age being a secondary consideration. This was because the forces applied to each patient-specific DEA model were scaled to the individual’s bodyweight, and it was undesirable to compare contact stresses between models loaded with vastly different forces.

Common radiographic measures of hip dysplasia (LCEA, Tönnis angle, extrusion index) normalized in these patients after PAO, but there are subtilties of deformity correction that may not be identified with these measurements. Anterior and posterior coverage are difficult to assess on plain films. It is possible that if each patient’s acetabulum was oriented in a patient-specific optimized position, contact stresses after PAO could be further reduced and may approach normal. However, given the significant differences between normal and dysplastic hips after PAO in this investigation, it is unlikely that these hips will ever achieve completely normal joint contact mechanics. A further limitation is that PAOs for all individuals selected for this study were performed 10-15 years ago (between 2003-2009), at a time when adjunctive procedures, such as corrections of femoral version and head-neck offset deformity were less common. While patients with dysplasia that were matched to the individuals with normal anatomy were only selected if they did not have obvious femoral deformity, the CT scans did not extend distally enough to evaluate femoral version. Femoral head-neck deformity was also not addressed at the time of PAO. Given that femoral deformities have been found to increase intra-articular contact stresses even during low-flexion activities such as gait,³⁹ it is possible that if correction of femoral deformity had also been performed in some of the dysplastic patients, the contact stress may have normalized further than what was achieved with PAO alone.

While PAO was not able to restore normal contact mechanics to dysplastic hips, it was able to substantially improve the mechanical environment of the joint. The locations of elevated contact stress in dysplastic hips were medialized, and the contact stress pattern during the stance phase of gait more closely resembled that of the normal hips after a PAO. These findings indicate that in addition to the improved joint stability and reduction in patient pain, there are meaningful improvements in intra-articular joint mechanics after PAO. However, the ongoing elevated joint contact stress after PAO may be a factor contributing to the degeneration of more than half of dysplastic joints treated with a PAO at follow-up times greater than 20-30 years.^{40, 41}