Dysplasia of the hip is a frequent cause of secondary osteoarthritis (OA) [29, 31, 32]. Although THA is the preferred treatment for older patients who are not expected to outlive the implant or in whom only few revisions are anticipated, arthroplasty becomes a less desirable solution in younger, more active patients in whom the cause of OA can be mechanically corrected. The goal of contemporary joint-preserving hip surgery is to correct these anatomic abnormalities ideally to prevent or at least retard the development of secondary OA.

In 1984, the Bernese periacetabular osteotomy (PAO) for treatment of developmental dysplasia of the hip (DDH) in adolescents and adults was introduced by one of the senior authors (RG) [10]. We have performed this procedure more than 1100 times at our institution and its use has spread worldwide. The Bernese PAO reportedly has favorable short-term and 10-year results [6, 8, 21, 30, 33, 40, 41, 47, 57], but the longer-term outcomes are unknown.

We raised the following questions: (1) what is the 20-year cumulative survivorship of hips that undergo the Bernese PAO?; (2) will the clinical scores and (3) the radiographic measures be maintained with time?; (4) do any demographic (age, gender, body mass index), clinical (hip pain, range of motion [ROM], walking ability, pain provocation tests), radiographic (osteoarthrosis, morphologic features of the femoral head, acetabular coverage and orientation), or surgery-related factors (concomitant intertrochanteric osteotomy [IO] or previous surgeries) predict a poor outcome? Finally, we compared our outcomes with the natural history and with the results of other osteotomies.

Using our patient records, we identified all patients of the series who had known conversion to a THA and the date of conversion. Contacting the remaining patients by phone, again the date of THA was recorded, allowing for calculation of the cumulative survivorship. We clinically and radiographically evaluated the patients with surviving joints at the outpatient clinic. These results for the survivor and the nonsurvivor groups were compared with the preoperative and 10-year followup status and the initial data were tested for predictive factors.

At followup, five patients (seven hips [9%]) were not available. One patient with bilateral osteotomies died during the observation period 6 years postoperatively from a cause unrelated to the osteotomy. Four patients (five hips [7%]) were lost to followup; two had an uneventful postoperative course with followup of 1.2 and 1.8 years, respectively; three hips with a followup of 10.6, 11.3, and 11.5 years had good to excellent clinical results (Merle d’Aubigné and Postel scores ranging from 16 to 18) without severe radiographic signs of OA (OA Grade 2 or greater according to Tönnis [55]) or conversion to a THA at final followup. The minimum followup of the remaining 58 patients (68 hips [91%]) was 19 years (mean, 20.4 ± 1.1 years; range, 19–23 years). This study was approved by the local Institutional Review Board.

The details of the operative technique have been described [10, 45]. In short, the technique is performed through a modified Smith-Petersen approach, and four periacetabular osteotomies and a controlled fracture are performed to completely mobilize the acetabulum from the innominate bone (Fig. 1). It allows extensive potential for acetabular reorientation with three rotational degrees of freedom, including medial and lateral displacement of the fragment. The posterior column of the true pelvis remains intact maintaining stability through an intact continuity of the pelvic ring allowing for minimal internal fixation of the fragment and aftercare with early partial weightbearing and immediate mobilization without the need for external fixation.

Fig. 1A–C The Bernese periacetabular osteotomy is performed through a modified Smith-Petersen approach, including an osteotomy of the anterior superior iliac spine to dissect the inguinal ligament and the adjacent muscles. (A) With four periacetabular osteotomies and a controlled fracture, the acetabulum is completely mobilized from the innominate bone. (B) For fixation of the reoriented fragment, three cortical screws are used. (C) The posterior column of the true pelvis remains intact maintaining stability through an intact continuity of the pelvic ring.

---

One of the authors (SDS) assessed the 58 patients at last followup with the Merle d’Aubigné and Postel score [27]. The Merle d’Aubigné and Postel score was graded as poor below 12, fair from 12 to 14, good from 15 to 17, and excellent with 18 points [27]. The anterior impingement and apprehension tests were assessed indicating a possible labral lesion anterosuperior or posteroinferior [18, 46, 52]. Additionally, the full goniometric ROM and gait were analyzed. All parameters were documented preoperatively and at 10 and 20 years postoperatively.

We tested normal distribution of all continuous parameters with the Kolmogorov-Smirnov test. Because none were normally distributed, we used only nonparametric tests. The cumulative survivorship was performed according to Kaplan and Meier [19] with the end point defined as conversion to THA or fusion of the hip. To detect differences between preoperative and postoperative radiographic values, we used the Wilcoxon signed rank test for continuous data and Fisher’s exact test for binominal data. To compare clinical and radiographic parameters preoperative and at 10- and 20-year followups, the Friedman test was applied. The Cox proportional hazards model was used to detect factors predicting poor outcome and to calculate the corresponding hazard ratios [7]. When a predictive factor with a nonnominal scale was found, differences in terms of the survivorship between the two groups were calculated using the log-rank test. When a survival analysis regarding preoperative OA grades was performed, the only Grade 3 OA according to Tönnis [55] (that was not lost-to-follow-up) was treated as Grade 2 OA.

Forty-one hips (60%) were preserved at 20 years, which corresponds to a cumulative Kaplan-Meier survivorship of 60.5% (range, 48.7%–72.2%; 95% confidence interval) (Figs. 2, 3). Twenty-six hips (38%) were converted to a THA and one (1%) had fusion at a mean of 11.7 ± 5.9 years (range, 0.9–19.3 years) after the index operation (Fig. 4).

Fig. 2 The Kaplan-Meier survival analysis is shown with the conversion to a THA or hip fusion as the end points. Values are expressed as cumulative survivorship with 95% confidence interval in parentheses for each 5-year interval. Circles indicate censored data.

---

Fig. 3A–G  Radiographs are shown of a 26-year-old woman with a dysplastic right hip with (A) an LCE angle of 16° and an AI of 12° without any signs of osteoarthritis (Tönnis osteoarthritis Grade 0 [55]). Initially she presented with limping and a negative impingement test. The mean Merle d’Aubigné and Postel score was 15. (B) The postoperative LCE angle was 32° and the AI −2°. At the (C) 10-year and the (D) 20-year followups, no osteoarthitic changes were seen. At last followup, the mean Merle d’Aubigné and Postel was 17, limping disappeared, and the impingement test remained negative. (E) In the corresponding false-profile [24] views, (F) the preoperative ACE angle of 39° increased to 45° postoperatively and (G) no osteoarthritic changes were seen at the 10-year followup. LCE = lateral center edge angle; AI = acetabular index; ACE = anterior center edge angle.

---

Fig. 4A–G Radiographs are shown of a 46-year-old women with a (A) dysplastic right hip with cranial acetabular retroversion, an LCE angle of 8°, an AI of 24°, and preoperative osteoarthritis grade of 1 according to Tönnis [55]. Initially she presented without limping and a negative impingement test. The mean Merle d’Aubigné and Postel score was 16 as a result of a decrease in pain status. (B) With the PAO, good coverage could be achieved with a postoperative relatively highly anteverted acetabular fragment and a resultant prominent posterior wall with an LCE angle of 30° and an AI angle of 4°. (C) The osteoarthritis showed marked progression after 5 years with end-stage osteoarthritis (Tönnis Grade 3), and (D) 10 years after the PAO, a THA with an acetabular reinforcement ring was performed. In the corresponding false-profile [24] views, the (E) preoperative ACE angle of 2° increased to (F) 25° postoperatively and showed (G) early deterioration with end-stage osteoarthritis at the 10-year followup. LCE = lateral center edge angle; AI = acetabular index; ACE = anterior center edge angle.

---

At 20 years, the mean Merle d’Aubigné and Postel score of all patients decreased (p = 0.004) in comparison to the 10-year value and was similar to (p = 0.095) the preoperative score, mainly related to an re-increase of the pain status. The 41 surviving hips (60%) had a mean Merle d’Aubigné and Postel score of 15.8 ± 2.1 (Table 3) at last followup. Of these, eight hips (20%) were graded as excellent, 25 (61%) as good, six (15%) as fair, and two (5%) as poor. The mean Merle d’Aubigné and Postel score increased (p < 0.001) from the preoperative status in comparison to the scores at 10 years (Table 3). Flexion (p < 0.001), internal (p < 0.001) and external rotation (p < 0.001), and abduction (p = 0.005) in the survivor group were decreased compared with the preoperative status (Table 3). The prevalence of limping decreased (p = 0.023) at the 20-year followup compared with the preoperative status. The prevalence of the anterior impingement test was significantly increased (p = 0.044) at last followup compared with the preoperative status. The prevalence of the apprehension test was not significantly different (p = 1.000) at the 20-year followup compared with the preoperative status (Table 3).

We observed no major changes in any of the radiographic parameters during the 20-year postoperative period except the OA score. In comparison to the preoperative radiographic values, there was an increase of the postoperative lateral center edge angle (LCE), the anterior center edge angle (ACE), the caudocranial/anterior and posterior coverage, and the intactness of Shenton’s line (Table 2). In contrast, the acetabular index (AI), the extrusion index, the crossover, and the posterior wall signs decreased (Table 2). We observed no differences for the ACM angle (p = 0.846). We observed no progression (p = 0.052) of the OA grade in the survivor group from a preoperative mean of 0.4 ± 0.6 (range, 0–2) to 0.6 ± 0.6 (0–2) during the first 10 years. However, the grade increased (p < 0.001) to 1.1 ± 1.0 (range, 0–3) at 20 years. We found decreased (p < 0.001) survivorship with increasing preoperative grade of OA (Fig. 5). Hips with OA scores of 0 or 1 preoperatively (n = 52) had a survivorship of 75%, whereas those with a preoperative OA score of 2 or 3 (n = 16) had a survivorship of 13% (Fig. 5).

Fig. 5 This flowchart shows the distribution of osteoarthritis (OA) progression and the failure rate depending on the preoperative osteoarthritis score of all 68 evaluated hips.

---

Dysplasia of the hip frequently leads to early asymptomatic and symptomatic hip degeneration. The goal of contemporary joint-preserving hip surgery is to correct the anatomic abnormalities leading to early degeneration ideally to prevent or at least retard the development of secondary OA. The goal of the Bernese PAO is to specifically correct deficient acetabular coverage in hips with developmental dysplasia to prevent secondary osteoarthrosis. We raised the following questions: (1) what is the 20-year cumulative survivorship of hips that had the Bernese PAO?; (2) will the clinical scores and (3) the radiographic measures be maintained with time?; (4) do any demographic (age, gender, body mass index), clinical (hip pain, ROM, walking ability, pain provocation tests), radiographic (osteoarthrosis, femoral head morphology, acetabular coverage and orientation), or surgery-related factors (concomitant intertrochanteric osteotomy [IO] or previous surgeries) predict a poor outcome? Finally, we compared our outcomes with the natural history and with the results of other osteotomies.

In additional to the lack of a control group, there are three other limitations to our study. First, clinical parameters of all patients were assessed by different observers at each followup. This is unavoidable for a longitudinal study design spanning more than 20 years. In the literature, substantial interobserver agreement and intraobserver agreement have been reported for the Merle d’Aubigné and Postel score [20], limp [4], and ROM, particularly flexion [14], internal [14, 62] and external [14] rotation, and abduction [14, 26]. Most of these parameters differed between the preoperative and 20-year followup status and therefore we believe these limitations do not influence the results of this study to a large extent. We identified no publication about the reliability and correlation with morphologic findings intraoperatively or in MRIs of the anterior impingement test. This should be considered when comparing the rates of positive anterior impingement tests at the different followups or interpreting the anterior impingement test as a predictive factor. Second, this series of patients reflects the learning curve and the first experiences of a new surgical technique that could be associated with a higher rate of complications [40, 47]. Third, there was considerable heterogeneity in terms of indications, varying degrees of dysplasia [44], OA grade, high number of previous operations, and additional underlying diseases. Although these variations likely reduced the overall 20-year survivorship results, these variations allow an analysis of factors predicting long-term survival.

The surgical technique underwent some minor modifications since the first case in April 1984 (Fig. 6). The supraacetabular osteotomy is now placed more cranially from the anterior-inferior to the anterior-superior iliac spine to preserve vascularity of the acetabular fragment [2] and to improve the hold of the Schanz screw for reorientation (Fig. 1). In addition, since the introduction of the femoroacetabular impingement concept, more emphasis was put on avoiding anterior and lateral overcorrection or retroversion, which could be associated with an unfavorable outcome [34]. For approximately 10 years now, an additional arthrotomy is performed routinely allowing for treatment of labral cysts, observation of range of internal rotation and flexion, and for correction of an aspherical portion of the femoral head-neck offset by osteochondroplasty to improve impingement-free ROM. Therefore, we believe the outcomes with the current technique (including the modifications) would be superior to those in our current study.

Fig. 6A–C  The first Bernese PAO was performed in 1984 in a 13-year-old girl with a (A) subluxated femoral head articulating in a very shallow and retroverted acetabulum after proximal femoral focal deficiency. (B) The PAO was performed with a concomitant valgus intertrochanteric osteotomy. Three years later, progressive posterior subluxation in flexion resulting from a deficient posterior coverage made a posterior shelfplasty necessary. (C) Twenty-three years later, she presented with a good clinical result (Merle d’Aubigné and Postel score, 15).

---

We identified six factors predicting poor outcome (Table 4). One was demographic, three were clinical, and two were radiographic factors. A substantial number of these factors were associated with an already advanced stage of joint degeneration at the time of surgery as could be proven for other types of acetabular reorientation procedures: the preoperative OA score [21, 28, 30, 36, 43, 58], a decreased preoperative Merle d’Aubigné and Postel score [58], a preoperative limp, and a positive anterior impingement test indicating a labral lesion [18, 46, 52]. In addition, advanced age [64] was a risk factor for early conversion to THA. The only one of these parameters positively influenced by surgery was the postoperative extrusion index. An undercorrected hip with an extrusion index greater than 20% is prone to develop end-stage OA during the postoperative course. An aspherical femoral head could not be correlated with failure according to these data [30]. Compared with the 10-year result of the same patient series reported earlier [47], the Merle d’Aubigné and Postel score, preoperative positive anterior impingement test and limp, and postoperative extrusion index were new predictive factors. We could not confirm a diminished ACE or an acetabular index less than 0 or greater than 10° predicted poor outcome.

A comparison with the current literature on this topic is difficult. We considered these comparisons in three ways: with the natural course of dysplastic hips, with other acetabular reorientation techniques, and with alternative surgical treatments.

Describing the natural history of hips with dysplasia, Hartofilakidis et al. [12] reported all referred dysplastic hips with a subluxation (ie, broken Shenton’s line) would undergo a THA by the age of 45. These authors would not, of course, be able to comment on asymptomatic patients with subluxation or those who were not symptomatic enough to warrant a referral; many of these patients might undergo THA at a later age or not at all. Others [59] suggest an “inevitability of disabling coxarthrosis” in patients with recognized subluxation. In our study cohort, 54% of all hips with preoperative subluxation (Severin Grade 4 and higher) did not have additional surgery at a mean age of 41 years with a mean Merle d’Aubigné score of 16 points. Murphy et al. [31] reported that all hips with an LCE angle less than 16° or an acetabular index greater than 15° ultimately would develop end-stage OA. In our symptomatic patient cohort, 54% with these radiographic characteristics had no additional surgery at the latest followup (mean age, 44 years). Given more than half of our patients with subluxation did not have additional surgery, we suggest the PAO provides outcomes superior to the natural history in symptomatic patients.

Good results have been reported with isolated femoral varus ostetotomies [17] or for patients treated with open or closed reduction [1, 25]; however, the patients reported in these series do not represent those in our series.

Despite the fact that our series represents the learning curve of a technically demanding procedure in an inhomogeneous patient group with various previous surgical attempts to achieve sufficient coverage and several concomitant IO, we believe the 20-year results of the first 75 hips are promising. Based on the predictive factors we identified, we suggest the ideal patient for this surgical procedure is young (younger than 30 years) with no or slight preoperative OA (OA score 0 or 1) and no severe hip pain. A positive anterior impingement test suggests an anterior labral lesion and therefore a worse prognosis. A major surgical difficulty is to find the correct balance between undercorrection and overcorrection of the acetabular fragment and to restore correct anteversion. Undercorrection should be avoided because a postoperative extrusion index less than 20% predicts worse outcome. However, overcorrection with acetabular retroversion may cause subsequent painful femoroacetabular impingement [34], which also predicts a worse outcome. We believe these long-term results show PAO is an effective technique for treating symptomatic DDH in selected patients and can maintain the natural hip for at least 19 years in most patients.