Search tips
Search criteria 


Logo of intorthopspringer.comThis journalToc AlertsOpen ChoiceSubmit Online
Int Orthop. 2009 December; 33(6): 1483–1488.
Published online 2008 August 22. doi:  10.1007/s00264-008-0643-7
PMCID: PMC2899154

Language: English | French

Factors predicting the failure of Bernese periacetabular osteotomy: a meta-regression analysis


There is no clear evidence regarding the outcome of Bernese periacetabular osteotomy (PAO) in different patient populations. We performed systematic meta-regression analysis of 23 eligible studies. There were 1,113 patients of which 61 patients had total hip arthroplasty (THA) (endpoint) as a result of failed Bernese PAO. Univariate analysis revealed significant correlation between THA and presence of grade 2/grade 3 arthritis, Merle de’Aubigne score (MDS), Harris hip score and Tonnis angle, change in lateral centre edge (LCE) angle, late proximal femoral osteotomies, and heterotrophic ossification (HO) resection. Multivariate analysis showed that the odds of having THA increases with grade 2/grade 3 osteoarthritis (3.36 times), joint penetration (3.12 times), low preoperative MDS (1.59 times), late PFO (1.59 times), presence of preoperative subluxation (1.22 times), previous hip operations (1.14 times), and concomitant PFO (1.09 times). In the absence of randomised controlled studies, the findings of this analysis can help the surgeon to make treatment decisions.


Le devenir de l’ostéotomie péri-acétabulaire Bernoise n’est pas évident. nous avons réalisé une méta analyse sur 23 études regroupant 1113 patients sur lesquels 61 patients ont nécessité du fait de l’échec de l’ostéotomie une prothèse totale. une analyse univariable montre une corrélation significative entre la prothèse et la présence d’une arthrose de grade 2 ou 3, le score de Merle d’Aubigne, le score de Harris, l’angle de Tonnis, les modifications de l’angle LCE, les ostéotomies fémorales tardives, la résection d’ossifications. L’analyse multivariable montre, qu’il y a 3,36 fois plus de chances d’avoir une arthrose s’il existe une arthrose de grade 2 ou 3, 12% s’il existe une pénétration articulaire, 1,09% si le score MDS est relativement bas, l’interpénétration articulaire 3,12 fois, le score de Merle d’Aubigné très bas 1,59 fois, une ostéotomie fémorale tardive 1,59 fois, une subluxation préopératoire 1,22 fois, une intervention préalable au niveau de la hanche 1,14 fois, une ostéotomie proximale fémorale concomitante 1,09 fois. Cette analyse permet au chirurgien de prendre des décisions thérapeutiques.


Ganz et al. initially described the Bernese periacetabular osteotomy (PAO) in 1988 on 75 hips with dysplasia [7]. In that series, only one patient needed conversion to total hip arthroplasty due to intra-articular joint penetration. Following this report, various authors have reported their experience with Bernese PAO. Most of these reports were retrospective case series with no control group for comparison. Lack of scientific evidence leads to significant controversies with respect to preoperative inclusion criteria. Various authors even considered patients with severe dyplasia, aspherical head, and significant underlying arthritis for PAO [2, 3, 20]. Very few studies attempted to define the factors that could predict the failure of periacetabular osteotomy. Siebenbrock et al., in their retrospective review of 63 patients, showed that age at surgery, presence and grade of osteoarthritis, presence of labral lesion, postoperative acetabular index above 10°, and postoperative anterior centre edge angle less than 40° can negatively influence the outcome of periacetabular osteotomy [19]. Cunningham et al., in their study involving 47 patients undergoing PAO, tried to determine the radiological parameters that would significantly affect the outcome [5]. They found that such factors as joint space width, joint subluxation, and preoperative osteoarthritis grade can negatively influence the outcome. However, they used revision to total hip arthroplasty (THA) and WOMAC scores as the indicators of failure as opposed to the Merle d’Aubigne score used by Siebenbrock et al. However, none of these studies had adequate patient populations or definite indicators of failure. A well conducted metanalysis may help better define the factors associated with failure of PAO. However, in the absence of any randomised controlled study on PAO, it was not possible to perform a meta-analysis. Hence, we performed a quantitative systematic analysis by pooling the data presented in all outcome studies published since 1988 and tried to determine the factors that would influence the failure of the procedure in terms of conversion to THA.

Methods and materials

Search strategy

In this study, one of the authors (SSN) initially conducted an extensive literature search to identify all articles reporting outcome following Bernese periacetabular osteotomy. We restricted our search to reports in the English literature. Our primary search was using Medline/Pubmed. We used MESH-term osteotomy with periacetabular (107 total, 15 reviews), Ganz (88 total, 17 reviews), Bernese (39 total, 7 reviews), pelvis (252 total,14 reviews), dysplasia (701 total, 63 reviews), and acetabular dysplasia (295 total, 31 reviews). We also used other search strategies like EMBASE and CINAHL using the OVID interface and searched other interfaces including EBSCO, Academic Search Premier and Proquest, as well as individual publishers’ websites such as MD consult, Science Direct, Lippincott Wilkins and Williams (LWW), Lippincott Ravens, Blackwell Synergy, Wiley Interscience, Taylor and Francis, and Springerlink. We used search terms such as “Bernese osteotomy”, “Ganz osteotomy”, “periacetabular osteotomy”, “pelvic osteotomy”, “acetabular dysplasia osteotomy” either alone or in different combinations. Our search identified 26 studies reporting outcome following Bernese periacetabular osteotomy. We retrieved the full text of all the articles to assess eligibility for the study.

Inclusion and exclusion criteria

We included all the studies with follow-up of at least 18 months. We included the studies with less follow-up or duration of follow-up not mentioned if they mentioned total hip arthroplasty in their results. Articles were excluded if they were review articles or letters. We excluded studies if there was duplication of data. We used the size of the study and study location to identify the redundant data.

Data extraction

Based on the above selection criteria, 23 studies were eligible for our analysis (Table 1). We considered the failure of PAO leading to total hip arthroplasty as the primary end point. Information regarding the study characteristics, preoperative clinical and radiological parameters, postoperative clinical and radiological parameters, intraoperative and postoperative complications and profile of patients undergoing total hip arthroplasty was collected wherever given (Table 2).

Table 1
Studies included in this review
Table 2
Variables studied in this review

Statistical analysis

Descriptive statistics were used to analyse the baseline variables. Pearson’s correlation, Spearman’s rho, and the chi-square test were used where appropriate for univariate analysis (Table (Table3).3). Those variables which showed significant correlation or association in the univariate analysis were included in the multivariate analysis. Multivariate analysis, using a binary logistic regression (meta-regression) model, was performed to assess the influence of heterogeneity of the studies included as well as to predict the explanatory variables.

Table 3
Univariate analysis of variables


Our study included 23 outcome reports with 1,113 patients with an average of 48 patients per study. Of these 1,113 patients, 61 patients had total hip arthroplasty as a result of failed Bernese periacetabular osteotomy.

Univariate analysis

Date of publishing

Studies published before 2000 reported significantly more THA (total = 42) as compared to the studies after 2000 (total = 19).


Average age of patients in the review was 30 (range 22.7–37.6). Average age of the patients who had conversion of PAO to THA was 30.9 as compared to 28.7 in those who had successful PAO. This difference was not statistically significant (p = 0.44). Also, there was no significant correlation between age and the number of THA (r = 0.19).

Neurogenic dysplasia and secondary dyplasia

Our analysis failed to show any statistically significant difference in the THA incidence in those studies which included neurogenic dysplasia (p = 0.06) and other secondary dysplasias (p = 0.33) as compared to those which did not include these patients. Similarly, there was no correlation between number of patients with either neurogenic dysplasia or other secondary dysplasia and the number of THA.

Number of previous operations

It was generally believed that outcome following PAO is less than satisfactory in patients who had other previous pelvic procedures to treat dysplasia. There were nine studies which mentioned inclusion of patients with previous pelvic surgery. However, our analysis failed to reveal any significant correlation between previous pelvic surgery and failure of PAO (r = 0.50).

Number of patients in the study and length of follow-up

Number of patients and duration of follow-up showed significant correlation with number of failures in the univariate analysis (r = 0.58 and 0.50, respectively). However, multivariate analysis failed to reveal any significant correlation.

Preoperative Merle de’Aubigne score and Harris hip score

Merle de’Aubigne score was used in seven studies and Harris hip score was used in eight studies. Average preoperative Merle de’Aubigne score was 13.25, and average preoperative Harris hip score was 64.5. Low preoperative Harris score failed to reveal any significant correlation to failure of PAO (r = −0.49), but low Merle de’Aubigne score showed significant correlation to failure of PAO in the univariate analysis (r = −0.78).

Radiological variables

Preoperative lateral centre edge (LCE) and anterior centre edge (ACE) and postoperative Tonnis angle, LCE, and ACE failed to show any significant correlation to failure of PAO. However, the change in the LCE angle following surgery was significantly more in patients with successful PAO as compared to those in whom PAO failed (p < 0.05).

Preoperative grade 2/grade 3 arthritis

Our study reveals significant correlation between the conversion to THA and inclusion of Tonnis grade 2 and 3 arthritis patients (r = 0.56).

Heterotrophic ossification (HO)

Eight studies mentioned postoperative formation of HO. Our analysis reveals no correlation between HO and failure of PAO (r = 0.51).

Inclusion of patients with subluxation and false acetabulum

Seven studies included patients with subluxation and false acetabulum. Our analysis revealed significant correlation of this variable to failure of PAO (r = 0.87).

Joint penetration

Intraoperative joint penetration during PAO is considered to be associated with poor outcome by most pelvic surgeons. In contrast, our univariate analysis failed to show any significant correlation between joint penetration and conversion to THA (r = 0.26).

Combined PAO with PFO and number of late proximal femoral osteotomies

We postulated that the need for proximal femoral osteotomy either at the time of PAO or at a later time suggests inclusion of a severely dysplastic hip and hence poor outcome. Our analysis revealed significant correlation between the failure of PAO and the need for late PFO (r = 0.66).

Multivariate analysis

We included all variables which were significant in univariate analysis for the multivariate analysis. We also included nonsignificant variables like secondary dysplasia, combined PAO with PFO, and joint penetration in the multivariate analysis to assess their ability to predict the conversion of PAO to THA. Three different logistic regression models were constructed based on the multicollinearity of the independent variables. None of the explanatory variables studied were able to significantly predict the chance of total hip arthroplasty in a patient undergoing PAO. However, it was evident from the regression model that the odds of having THA increases by 3.36 times with the presence of preoperative grade 2 and grade 3 osteoarthritis, 3.12 times with intraoperative joint penetration, 1.59 times with low preoperative Merle de’Aubigne score, 1.59 times with late PFO, 1.22 with presence of preoperative subluxation, 1.14 times with previous hip surgeries, and 1.09 times with concomitant PFO.


Our study shows the importance of various preoperative variables in predicting the failure of PAO. The biggest outcome series to date was reported by Murphy et al. in 1999. In their series involving 130 hips, only five failed [16]. He included patients with grade 2 and grade 3 osteoarthritis in his study, but he failed to mention in his report the proportions of the hips which had advanced osteoarthritis. Further, he used hip score as an outcome variable instead of THA. Siebenbrock et al., in their series of 75 patients, reported 13 THAs [19]; however, they failed to assess the factors leading to the failure of PAO.

Severity of preoperative osteoarthritis and intraoperative joint penetration have been consistently shown by various authors to be associated with poor outcome of PAO [5, 7, 8, 11, 19]. Our analysis further confirms this fact and shows that the odds of THA increases by 3.36 times with severe preoperative OA and 3.12 times with intraoperative joint penetration. Hence, on the basis of current available evidence and our analysis, we would recommend that surgeons be cautious when recommending Bernese PAO to patients with dysplasia and severe secondary OA changes. Further, surgeons performing PAO should take all measures including cadaveric training to understand the complex 3D orientation of the pelvis, preoperative computerised tomographic scan of each patient, and intraoperative image guidance to avoid intraoperative joint penetration.

Our analysis shows no significant correlation between neurogenic dysplasia and failure of Bernese PAO. This is similar to an exclusive report by MacDonald et al. of Bernese PAO in 13 patients with neurogenic dysplasia [12]. It is a commonly held belief among pelvic surgeons that previous pelvic and hip surgery could adversely affect the outcome of PAO. However, Mayo et al., in their report of 19 PAO patients with previous pelvic surgeries, failed to reveal any such correlation [14]. Similarly, our study failed to show any significant association between previous pelvic operations and failure of PAO in both univariate and multivariate analysis.

In our analysis, we assessed the impact of severity of dysplasia on the outcome of THA. At the start of the analysis, we postulated that severe dysplasia (Severin grade 4 [subluxation] and grade 5 [false acetabulum]) can lead to poor outcome and failure of PAO. Review of the literature revealed that only seven studies have mentioned the preoperative severity of dysplasia, and only four studies mentioned the exact number of patients with severe dysplasia. However, none of these studies exclusively reported the outcome of PAO in patients with severe dysplasia. Most studies failed to report the proportion of patients with severe dysplasia in their study population. Hence, we considered other variables such as need for concomitant PFO or need for PFO at a later date as indicators of severe dysplasia. Our analysis reveals that odds of THA increases by 22% with subluxation, 9% with the need for concomitant PFO, and 54% with late PFO.

We understand that the weight of this study is very limited as there were not any randomised controlled studies on Bernese periacetabular osteotomy in the literature. Further, we included a heterogeneous group of studies in our analysis. Hence we are wary about jumping to any conclusions based on our analysis. However, it has to be agreed that in the absence of randomised controlled trials and well-conducted multicentre studies and limited but confounding evidence, our systematic analysis should help to guide the surgeons and patients in their decision making process.


1. Armand M, et al. Outcome of periacetabular osteotomy: joint contact pressure calculation using standing AP radiographs, 12 patients followed for average 2 years. Acta Orthop. 2005;76(3):303–313. [PMC free article] [PubMed]
2. Clohisy JC, et al. Periacetabular osteotomy for the treatment of severe acetabular dysplasia. J Bone Joint Surg Am. 2005;87(2):254–259. doi: 10.2106/JBJS.D.02093. [PubMed] [Cross Ref]
3. Clohisy JC, et al. Periacetabular osteotomy for the treatment of acetabular dysplasia associated with major aspherical femoral head deformities. J Bone Joint Surg Am. 2007;89(7):1417–1423. doi: 10.2106/JBJS.F.00493. [PubMed] [Cross Ref]
4. Crockarell J, Jr, et al. Early experience and results with the periacetabular osteotomy. The Mayo Clinic experience. Clin Orthop Relat Res. 1999;363:45–53. doi: 10.1097/00003086-199906000-00007. [PubMed] [Cross Ref]
5. Cunningham T, et al. Delayed gadolinium-enhanced magnetic resonance imaging of cartilage to predict early failure of Bernese periacetabular osteotomy for hip dysplasia. J Bone Joint Surg Am. 2006;88(7):1540–1548. doi: 10.2106/JBJS.E.00572. [PubMed] [Cross Ref]
6. Davey JP, Santore RF. Complications of periacetabular osteotomy. Clin Orthop Relat Res. 1999;363:33–37. doi: 10.1097/00003086-199906000-00005. [PubMed] [Cross Ref]
7. Ganz R, et al. A new periacetabular osteotomy for the treatment of hip dysplasias. Technique and preliminary results. Clin Orthop Relat Res. 1988;232:26–36. [PubMed]
8. Garras DN, et al. Medium-term results of the Bernese periacetabular osteotomy in the treatment of symptomatic developmental dysplasia of the hip. J Bone Joint Surg Br. 2007;89(6):721–724. doi: 10.1302/0301-620X.89B6.18805. [PubMed] [Cross Ref]
9. Hsieh PH, et al. Image-guided periacetabular osteotomy: computer-assisted navigation compared with the conventional technique: a randomized study of 36 patients followed for 2 years. Acta Orthop. 2006;77(4):591–597. doi: 10.1080/17453670610012656. [PubMed] [Cross Ref]
10. Katz DA, et al. Periacetabular osteotomy in patients with Down’s syndrome. J Bone Joint Surg Br. 2005;87(4):544–547. doi: 10.1302/0301-620X.87B4.15314. [PubMed] [Cross Ref]
11. Kralj M, et al. The Bernese periacetabular osteotomy: clinical, radiographic and mechanical 7–15-year follow-up of 26 hips. Acta Orthop. 2005;76(6):833–840. doi: 10.1080/17453670510045453. [PubMed] [Cross Ref]
12. MacDonald SJ, et al. Periacetabular osteotomy in the treatment of neurogenic acetabular dysplasia. J Bone Joint Surg Br. 1999;81(6):975–978. doi: 10.1302/0301-620X.81B6.9700. [PubMed] [Cross Ref]
13. Matta JM, et al. Periacetabular osteotomy through the Smith-Petersen approach. Clin Orthop Relat Res. 1999;363:21–32. [PubMed]
14. Mayo KA, et al. Results of periacetabular osteotomy in patients with previous surgery for hip dysplasia. Clin Orthop Relat Res. 1999;363:73–80. doi: 10.1097/00003086-199906000-00010. [PubMed] [Cross Ref]
15. Murphy SB, Millis MB. Periacetabular osteotomy without abductor dissection using direct anterior exposure. Clin Orthop Relat Res. 1999;364:92–98. doi: 10.1097/00003086-199907000-00013. [PubMed] [Cross Ref]
16. Murphy SB, et al. Surgical correction of acetabular dysplasia in the adult. A Boston experience. Clin Orthop Relat Res. 1999;363:38–44. doi: 10.1097/00003086-199906000-00006. [PubMed] [Cross Ref]
17. Peters CL, et al. Early results of the Bernese periacetabular osteotomy: the learning curve at an academic medical center. J Bone Joint Surg Am. 2006;88(9):1920–1926. doi: 10.2106/JBJS.E.00515. [PubMed] [Cross Ref]
18. Pogliacomi F, et al. Periacetabular osteotomy. Good pain relief in symptomatic hip dysplasia, 32 patients followed for 4 years. Acta Orthop. 2005;76(1):67–74. doi: 10.1080/00016470510030346. [PubMed] [Cross Ref]
19. Siebenrock KA, et al. Bernese periacetabular osteotomy. Clin Orthop Relat Res. 1999;363:9–20. doi: 10.1097/00003086-199906000-00003. [PubMed] [Cross Ref]
20. Trousdale RT, et al. Periacetabular and intertrochanteric osteotomy for the treatment of osteoarthrosis in dysplastic hips. J Bone Joint Surg Am. 1995;77(1):73–85. [PubMed]
21. Trousdale RT, et al. Magnetic resonance imaging pelvimetry before and after a periacetabular osteotomy. J Bone Joint Surg Am. 2002;84-A(4):552–556. [PubMed]
22. Trumble SJ, et al. The periacetabular osteotomy. Minimum 2 year followup in more than 100 hips. Clin Orthop Relat Res. 1999;363:54–63. doi: 10.1097/00003086-199906000-00008. [PubMed] [Cross Ref]
23. Bergayk AB, Garbuz DS. Quality of life and sports-specific outcomes after Bernese periacetabular osteotomy. J Bone Joint Surg Br. 2002;84(3):339–343. doi: 10.1302/0301-620X.84B3.12421. [PubMed] [Cross Ref]

Articles from International Orthopaedics are provided here courtesy of Springer-Verlag