In this paper we present an extended follow-up of 12 patients who underwent periacetabular osteotomy. In a previous publication (Armand et al. 2005
), we presented a 2-year follow-up of these patients including clinical, radiological, and 2-dimensional mechanical evaluations. The present study implements improved mechanical and radiological analysis in addition to a longer-term (10-year) clinical follow-up. We have found no previous studies using 3D mechanical analysis for PAO on both pre- and post-operative CT data.
Comparison of the results of this 3D mechanical analysis to those of the previous 2D analysis of the same patients showed some obvious improvements in the modeling technique. The magnitudes of peak pressure calculated using the 3D contact surface ranged from 1.9 to 7.7 MPa in the preoperative cases and from 1.4 to 3.2 MPa postoperatively. These peak pressure values have the same magnitude as those reported in similar 3D contact studies of the hip (Tsumura et al. 2005
, Yoshida et al. 2006
). Using only the 2D geometry, a much wider and unrealistic range of pressures (3–128 MPa preoperatively and 2–34 MPa postoperatively) was found in the same 12 patients (Armand et al. 2005
). Secondly, in the previous 2D study, centroid of pressure (CP) ratios within the range 0.4–0.6 were considered to be ideal; however, optimization of the entire 3D geometry of the hip resulted in an optimal range of CP between 0.6 and 0.85 (slightly medial).
Following surgery, the PAO reduced peak pressure in all but 1 case, while lateral coverage increased in every patient. Over-correction (increasing the lateral coverage beyond the normal CE range of 33 (SD 10) degrees (Janzen et al. 1998
) results in negative AC angles and peak pressure located at the medial aspect of the joint (Armand et al. 2005
). In this 3D study, the mechanical consequences of overcorrection were not as severe as shown in the 2D study. Only patient 9 showed an increase in pressure, by 5%. The anterior and posterior horns of the contact surface, not seen in the 2D model, contributed to the final mechanical outcome to mitigate an exponential rise in pressure with increased abduction of the acetabular cup. This suggests that the most important consequence of overcorrection of the acetabular fragment orientation is not the peak pressure, but a different factor not included in this model (e.g. impingement). An exponential rise in pressures corresponded to reduced lateral coverage of the femoral head, most likely due to the exclusion of the soft tissue labrum in this model.
An et al. (1990)
discussed 3 physiological cases in the context of the mechanical model regarding the resultant force vector (R) of loading through the joint and the “virtual displacement” vector (U): (1) a stable joint is one in which both R and U pass through the articular surface; (2) joint subluxation is indicated when only R passes through the articular surface and the vector U is outside the joint space; and (3) joint dislocation occurs when both R and U are directed outside the articular surface. In the present investigation, the preoperative analyses showed 11 of the 12 patients with the U vector directed outside and lateral to the articular surface (i.e. at risk of subluxation). Postoperative analysis showed all 12 hips corrected to stable joint conditions according to the hypothesized “virtual displacement” criteria. The angle between these 2 vectors (RU angle) is an additional mechanical parameter that indicates the quality of the joint alignment.
For optimization, the objective was to minimize the peak pressure. Previous studies have shown that a minimum peak pressure exists by varying the orientation of the acetabular fragment (Tsumura et al. 2005
). The contact area parameter was not sensitive enough to be a basis for planning PAO through optimization. Contact area increased in all 12 PAO cases, despite an increase in peak pressure and overcorrection of the radiological angles in one patient. The optimization routine minimizing peak contact pressure produced stable results and was evident because of the collinear (less than 1 degree) vector orientations of the peak pressure and virtual displacement. Compared radiologically, the optimized cases had CE angles in the normal range and the AC angles ranged from 1 to 12 degrees, which is consistent with the PAO surgical plan. The optimized CP ratio in this 3D study was greater (more medial) than that of the previous 2D mechanical analysis owing to the inclusion of the anterior and posterior horns in the complete 3D model. The optimization results were consistent with normal anatomy and the traditional preoperative aims of PAO. Planning of accurate correction in acetabular reorientation on 2D radiographs seems to pose limitations. Our results suggest that 3D evaluation of mechanical conditions greatly improves preoperative planning and has the potential to improve finding the right orientation for joint surfaces during surgery.
A limiting factor of this mechanical model is the calculation of the joint force, which predicates the pressure profile obtained using the DEA technique. Other optimization routines minimize pressure for only a single joint force load (Tsumura et al. 2005
) while this study optimized for 3 different joint forces simultaneously. However, the in vivo dataset of joint forces (measured with an instrumented endoprosthesis (Heller et al. 2001
)) has an unknown correlation to dysplastic patients, since that group included normal patients that had undergone hip arthroplasty. In addition, to apply these data correctly, the joint force must be applied using the same coordinate system as defined when recording data. The L5-S1 junction, used as a landmark in defining the local pelvis coordinate system, is not usually visible in preoperative PAO CT scans. Because patients are scanned in supine (non-weight-bearing) position, an accurate frame for mechanical analysis is not available and must be assumed.
Interobserver variability was greater in the preoperative cases than in the postoperative and optimized cases. The main factor contributing to this effect is discrepancies between observers during segmentation of the joint contact region. Variance during segmentation can affect the shape of the contact surface and this in turn can lead to variance within subsequent mechanical analyses. The effect is amplified when the joint is in an unstable condition due to dysplasia (preoperative) whereby changes to the joint surface result in large changes in contact pressure. The effects of interobserver variability are less pronounced when the joint is stable in either the postoperative and optimum configuration.
The DEA technique has been experimentally evaluated in cadavers, focusing on trends rather than absolute pressure magnitudes (Elias et al. 2004
). Validation studies in dysplastic hips are not feasible due to the limited availability of appropriate specimens, and limitations of experimental models. To offset the effect of unknown absolute correlation, the mechanical results presented here reflect the ratio of pressure, rather than the absolute magnitude of pressure results, in the pre- to postoperative comparisons. A limitation associated with the use of CT data was that the cartilage tissue was not directly visible, and hence this model involved segmenting the sub-chondral bone of the pelvis and used uniform cartilage thickness, though thickness varies in both normal and dysplastic hips (Nishii et al. 2004
Comparison of mechanical and radiological results to clinical outcomes directly using statistical methods would require more patients. Little variability existed between the 12 cases, 11 of which had excellent Harris score results. 1 patient with a poor outcome had radiological evidence of disruption of the anterior aspect of the load-bearing surface of the joint following PAO, which may have contributed to the low score. Studies on more cases with the present 3D model are needed in order to better understand the correlation between clinical outcome and the mechanical parameters described. Based on the mechanical model, we expected the change in contact pressure due to PAO to be related to patient outcome q-scores. That is, achieving the minimum contact pressure should give a favorable outcome whereas higher postoperative pressures would result in less favorable outcomes. Statistical power analysis based on the data characteristics observed in this study showed that 50 cases would be required to demonstrate significant correlation between contact pressure and patient outcome for a power of 0.8 and p < 0.05.
To conclude, the 3D mechanical model offers improvements over previous 2D models by including the anterior and posterior horns of the acetabular contact surface. This investigation indicates that optimization of the peak pressure criteria produces results that are consistent with normal patient anatomy and current surgical goals for the procedure. The rapid computation time of DEA and the radiological measurement methods make these tools potential applications in invoking real-time assessment tools for surgery.