Introduction

Retroversion of the acetabulum is an accepted cause of impaction-type femoroacetabular impingement as a result of pincer-type pathomorphology [35], which reportedly is associated with hip pain and early osteoarthritis in young adults [9–11, 53]. The anterior overcoverage of the hip caused by the acetabular retroversion leads to early, painful osseous contact in hip flexion and internal rotation and subsequently to localized labrum and cartilage damage.

Acetabular retroversion was initially understood as a prominent overgrowth of the anterior acetabular wall [4], but there is increasing evidence that acetabular retroversion is a rotational abnormality of the entire hemipelvis [15, 16, 28, 36, 44, 49] rather than focal overgrowth.

In the last two decades, several approaches have been used to address anterior overcoverage. These have consisted of localized débridement or removal of the anterior acetabular rim via either open or arthroscopic techniques [29, 30, 33, 34, 43, 57] or by reorientation of the entire acetabulum by periacetabular osteotomy (PAO) resulting in anteversion [31, 39, 40]. At short-term followup (defined as 2–4 years), both surgical treatments have demonstrated substantial improvements in clinical scores without radiographic progression of osteoarthritis [29–31, 33, 34, 39, 40, 43].

However, in the past several years, in our own practice, we noted that patients treated with anteverting PAO appeared to demonstrate superior results over the longer term to those treated with rim trimming. We posited that this observation supported the idea that rotational malalignment is the main source of acetabular retroversion, because the shape of the outer margin of the acetabular rim and the size of the lunate surface are essentially normal [23, 44].

Based on this background, we asked if anteverting PAO results in an increased cumulative 10-year survivorship of the hip compared with rim trimming through a surgical hip dislocation in a comparable group of patients with symptomatic acetabular retroversion (endpoints defined as a Merle d’Aubigné score of < 15 points [6]; radiographic progression of osteoarthritis greater than one Tönnis grade [52]; and/or conversion to THA).

Material and Methods

We performed a retrospective, comparative study comparing the midterm survivorship of symptomatic hips with acetabular retroversion treated either with acetabular rim trimming or anteverting PAO. Based on our institutional database of surgical hip dislocations (1436 hips) and PAOs (520 hips) between 1997 and 2012 (Fig. 1), we identified all hips that had surgery for acetabular retroversion (279 hips) at our institution (Fig. 2). Acetabular retroversion was defined as a concomitantly present positive crossover sign of the anterior and posterior acetabular rim, a posterior wall sign, and ischial spine sign [16, 35].

Fig. 1
figure 1

The algorithm for our patient selection is shown. LCE = lateral center-edge.

Full size image
Fig. 2
figure 2

The total number of patients undergoing PAO and surgical hip dislocation (SHD) and the number of patients included in this study between 1997 and 2012 are shown.

Full size image

During the period in question, the choice of surgical technique was based on the most recent understanding of the pathology and the availability of the surgical technique. The indication was consistently painfully restricted hip ROM in flexion and internal rotation with typical reproduction of groin pain. Before 2004, we generally treated acetabular retroversion by anteverting PAO (Fig. 2). Surgical hip dislocations have been performed since 1992 in our institution mainly for treatment of cam-type deformities [8]. Open rim trimming was introduced at the end of the 1990s. Labral refixation was introduced in 2001 and with growing experience and early success of rim trimming in less distinct acetabular retroversions, we applied this technique in hips with more pronounced acetabular retroversion with all three radiographic signs (crossover/posterior wall/ischial spine signs). This explains the decreased number of anteverting PAOs and the increased number of open acetabular rim trimmings around Year 2004 (Fig. 2). The number of surgical hip dislocations with rim trimming reached the maximum between 2004 and 2008. During this time, we learned that acetabular retroversion (defined as having all three signs) is a result of rotation of the hemipelvis [8, 15, 16, 44, 49]. In addition, the size of the lunate surface and the outer margin of the acetabular rim are normal in retroversion [44] and can theoretically be reduced in a critical way by rim trimming [3]. This explains the subsequent increased number of anteverting PAOs between 2007 and 2012 and is now our surgical treatment of choice for acetabular retroversion in most cases. During this transition time in Years 2007 to 2010, anteverting PAO was performed in younger patients, those younger than 35 years of age. Open acetabular rim trimming during this time period was indicated in patients older than 35 years of age with advanced anterosuperior cartilage damage, which would be rotated into the main weightbearing zone with anteverting PAO.

Because the use of the two surgical techniques was somewhat semisequential and overlapping resulting in different followup intervals, we sought to retrospectively look at all our procedures performed in our institution for acetabular retroversion, match the two groups, and finally compare the outcome with statistical methods that are specifically designed to account for variation in followup.

From those 279 hips treated for acetabular retroversion, we further excluded 57 hips with concomitant hip dysplasia (defined as a lateral center-edge angle of < 20° [27, 53, 54]; Fig. 1) and hips with incomplete radiographic information (23 hips). Of the remaining 199 hips (99 hips with anteverting PAO and 100 hips with rim trimming), 22 hips with an anteverting PAO were further excluded as a result of the following reasons: living abroad (three hips), complex morphologic abnormalities (eg, Legg-Calvé-Perthes disease, four hips), pregnancy (two hips), routine followup declined (six hips), and lost to followup (seven hips). Of the remaining 100 hips with rim trimming, the following 33 hips were excluded: living abroad (five hips), complex morphologic deformities (eg, Legg-Calvé-Perthes disease, two hips), death unrelated to surgery (one hip), routine followup declined (eight hips), and lost to followup (17 hips). The remaining 77 anteverting PAOs and 67 hips with surgical rim trimming were comparable regarding demographics and preoperative radiography except for more men and older patients in the rim trimming through surgical hip dislocation group. To account for this difference, the two groups were matched as follows: all patients for both groups were sorted for age and sex with blinding of the followup. Stepwise exclusion of the youngest woman of the anteverting PAO group and of the oldest man of the rim trimming through surgical hip dislocation group was performed unless the differences in age and sex reached a p value > 0.05. This led to an exclusion of 20 hips (10 for each group), eventually resulting in two equivalent groups of 67 hips with anteverting PAO and 57 hips with acetabular rim trimming through a surgical hip dislocation (Fig. 1). Independent from the group allocation, the indication for surgery was painfully restricted hip ROM in flexion and internal rotation with typical reproduction of groin pain.

To demonstrate that the two groups were highly comparable, we assessed demographic parameters, pre- and postoperative clinical parameters, and pre- and postoperative radiographic parameters. Four preoperative demographic factors were assessed (age, sex, side, body mass index) and these did not differ between the two groups (Table 1).

Table 1 Demographic and radiographic data of the two study groups
Full size table

Clinical parameters included the Merle d’Aubigné score [6], goniometric ROM (flexion/extension, internal/external rotation in 90° of flexion, ab-/adduction), and the presence of a positive anterior impingement test (ie, painful internal rotation in 90° of flexion), which were assessed by different observers because the study period covered more than one decade. We used the Merle d’Aubigné score [6], because this was the only consistently used clinical scoring system pre- and postoperatively. There were no differences between the two study groups for preoperative Merle d’Aubigné scores, clinical ROM of any amplitude, and the prevalence of a preoperative positive anterior impingement test (Table 1). The radiographic evaluation consisted of a preoperative and postoperative AP pelvic radiograph and a cross-table axial radiograph according to a standardized technique [50], which were available for all patients included in the final analysis. One observer (CAZ, not involved in the surgical care of the patients) assessed eight radiographic parameters preoperatively and postoperatively using previously developed and validated software, Hip2Norm (University of Bern, Bern, Switzerland) [47, 51, 56]. Osteoarthritis was graded according to Tönnis [52], which reportedly has good inter- and intraobserver variability [45]. Preoperatively, the two groups did not differ in terms of preoperative radiographic Tönnis grade of osteoarthritis [52], lateral center-edge angle, acetabular index, extrusion index, posterior acetabular coverage, or alpha angle (Table 2). The anteverting PAO group had slightly higher preoperative total acetabular coverage and higher anterior acetabular coverage [46]. Postoperatively, all evaluated radiographic values were comparable except for the posterior acetabular coverage and the retroversion index. This difference resulted inherently as part of the surgical procedures, because anterior coverage is reduced with simultaneous increase of the posterior coverage in a reorienting osteotomy, whereas only anterior coverage is reduced in rim trimming. In addition, one of us (MSH, not involved in the surgical care of the patients) reviewed the preoperative MR images to objectively assess the preoperative joint degeneration. Only MRIs with intraarticular gadolinium contrast and a standardized technique including radial sequences were reviewed [22]. These were available for 37 of the 67 hips (55%) of the anteverting PAO group and 48 of the 57 hips (84%) of the rim trimming through surgical hip dislocation group. Lesions of the labrum and the acetabular/femoral cartilage were assessed and graded according to a previously published standardized grading system [38], which had moderate interobserver agreement for labral lesions and substantial agreement for lesions of the acetabular/femoral cartilage. There were no differences between the two groups for preoperative labral and/or acetabular/femoral cartilage lesions (Table 1).

Table 2 Radiographic parameters of the two study groups pre- and postoperatively
Full size table

Anteverting PAO was performed according to a previously described technique (Fig. 3A–B) [40, 41]. None of the 67 hips had refixation of the labrum. Thirty-four patients (51% of the hips) underwent concomitant femoral osteochondroplasty through an anterior capsulotomy. Surgical hip dislocation with rim trimming was performed using a step-cut trochanteric osteotomy as described earlier (Fig. 3C) [8]. In all 57 hips (100%), labral refixation and femoral osteochondroplasty were performed. The rehabilitation protocol for both procedures consisted of early mobilization at postoperative Day 1, the use of continuous passive ROM, and partial weightbearing of 15 to 20 kg for 8 weeks for anteverting PAOs and 30 kg for 6 weeks for surgical hip dislocations.

Fig. 3A–C
figure 3

(A) The definition of acetabular retroversion is illustrated. We defined acetabular retroversion as presence of all three following signs: a crossover sign (asterisk) of the anterior (AW) and posterior (PW) acetabular wall, a positive ischial spine sign (intrapelvinic protrusio of the ischial spine, arrow), and a positive posterior wall sign (the posterior wall runs medial to the femoral head center). (B) The treatment with an anteverting PAO with optional concomitant femoral osteochondroplasty is shown. The labrum is typically not addressed. (C) The treatment by acetabular rim trimming with labral refixation through a surgical hip dislocation is shown. Two trochanteric screws are needed to refix the greater trochanteric osteotomy.

Full size image

Regardless of type of surgical treatment, patients were seen at 6 to 8 weeks, 3 months, 1 year, 2 years, and at 5-year intervals thereafter or in cases of an unsatisfactory postoperative course at any time. A minimum followup of 2 years was required for this study. Failures were included at any time. The median followup for the anteverting PAO group was 9.5 years (mean 9.1 ± 5.3 years; range, 2–17.4 years) and 6.8 years (mean 6.7 ± 2.3 years; range, 2.2–10.5 years) for the rim trimming through surgical hip dislocation group (p < 0.001).

We performed a power analysis for the primary research question regarding survivorship at 10 years followup with a two-sided level of significance of 5%, beta error of 5%, known long-term survivorship of hips undergoing anteverting PAO for acetabular retroversion of 71% with a SD of 17% [39], and known long-term survivorship of hips undergoing surgical hip dislocation of 80% for femoroacetabular impingement (FAI) with a SD of 5% [42] resulting in a minimal sample size of at least 35 hips per group.

We tested normal distribution with the Kolmogorov-Smirnov test. Normally distributed data between the two study groups were compared using the unpaired t-test; data that were not normally distributed were compared using the Mann-Whitney U-test. Normally distributed data between the pre- and postoperative status were compared using the paired t-test; data that were not normally distributed were compared using the Wilcoxon rank-sum test. Fisher’s exact test was used to compare binominal data among study groups. The survivorship was calculated using the method according to Kaplan and Meier [17] with the following endpoints: conversion to THA, radiographic progression of osteoarthritis greater than one grade according to Tönnis [52], and/or a Merle d’Aubigné-Postel score of < 15 points (so-called fair or poor) [6]. The log-rank test was used to compare the two survivorship curves.

Results

At 5 years, rates of clinical failure were the same for the two patient groups (86% [95% confidence interval {CI}, 76%–94%] for anteverting PAO versus 86% [95% CI, 76%–96%] for acetabular rim trimming). However, by 10 years, the patients treated with rim débridement fared worse with better survivorship at 10 years in hips undergoing anteverting PAO for acetabular retroversion (79% [95% CI, 68%–90%]) compared with acetabular rim trimming (23% [95% CI, 6%–40%]) at 10 years (p < 0.001; Fig. 4). The substantial drop for the rim trimming through surgical hip dislocation group started 6 years postoperatively. Analyzing the failures in the first 10 years, hips undergoing acetabular rim trimming presented mainly with a higher prevalence of a Merle d’Aubigné-Postel score of < 15 points (p = 0.007; Table 3), whereas we found no difference for the proportion that underwent conversion to THA (p = 0.908) or in radiographic progression of osteoarthritis (p = 0.953).

Fig. 4
figure 4

This figure shows the Kaplan-Meier survivorship for anteverting PAO for symptomatic acetabular retroversion compared with rim trimming through a surgical hip dislocation (SHD). Endpoints were defined as: conversion to THA, radiographic progression of osteoarthritis, and a Merle d’Aubigné-Postel score [6] of < 15 points. Although there was comparable survivorship at 5 years, the survivorship curve for hips undergoing acetabular rim trimming dropped at 6 years compared with anteverting PAO (p = 0.0002).

Full size image
Table 3 Reasons for failure for the two groups at 10 years
Full size table

Discussion

Acetabular retroversion can cause impaction-type FAI leading to hip pain and osteoarthritis. It can be treated by anteverting PAO (Fig. 5A–C) [31, 40, 41, 48] or acetabular rim trimming [2, 7, 13, 29, 42] with refixation of the labrum (Fig. 6A–C). In the past years, both techniques were in common use with similar short-term results. There is increasing evidence that acetabular retroversion is a rotational abnormality of the entire hemipelvis and not a focal overgrowth of the anterior acetabular wall, which would be in favor of an anteverting PAO. However, it is unknown if this larger procedure would be beneficial in terms of survivorship and Merle d’Aubigné scores at midterm followup compared with rim trimming. We evaluated the midterm followup of two comparable groups of patients with symptomatic acetabular retroversion undergoing either anteverting PAO or rim trimming through a surgical hip dislocation. Patients with acetabular rim trimming had comparable survivorship at 5 years but this decreased significantly after Year 6 (Fig. 4). The most commonly found reason for failure was a decreased Merle d’Aubigné score.

Fig. 5A–C
figure 5

An illustrative case of a 17-year-old patient with symptomatic acetabular retroversion is shown. (A) The preoperative image shows a positive crossover sign, a positive ischial spine sign, and a positive posterior wall sign. (B) The patient underwent anteverting PAO without radiographic signs of osteoarthritis at a 1-year (B) and 12-year followup (C). AW = anterior wall; PW = posterior wall.

Full size image
Fig. 6A–C
figure 6

An illustrative case of a 37-year-old patient with symptomatic acetabular retroversion is shown. (A) The preoperative morphology is very similar to the patient in Figure 5. All three signs for acetabular retroversion are positive. (B) The 1-year followup was favorable without signs of osteoarthritis. (C) However, the patient developed joint-space narrowing (circle) and osteophyte formation (arrow) at followup of 7 years indicating early OA. AW = anterior wall; PW = posterior wall.

Full size image

This study has limitations. First, our retrospective analysis does not imply a randomization of two procedures, which would be the ideal study design. The indication for either treatment has changed over time, which resulted in a semisequential series of two treatments. We note, however, that there were no clear, consistent indications for use of these procedures during the entire study period; rather, our surgical decision-making changed with time as we gained experience and understanding of acetabular pathology. These changes included the unavailability of the surgical technique of hip dislocation with trimming and labral refixation before 2004 and the improved scientific understanding of the underlying pathomorphology in acetabular retroversion around Years 2007 and 2012, which finally resulted in our chronologic distribution of these two surgeries in our department (Fig. 2). However, to account for our concerns for selection bias, we pooled all available patients from our institutional database with acetabular retroversion and checked for differences in demographics and preoperative morphology. Because the two groups differed in terms of sex and age, we finally matched the two groups of surgical treatment blinded to their followup by stepwise exclusion until differences in age and sex reached a p value < 0.05. This resulted in two highly comparable groups of patients regarding demographics, preoperative ROM, preoperative Merle d’Aubigné scores, amount of preoperative articular damage based on MRI, preoperative radiographic acetabular and femoral morphology, and accuracy of postoperative correction underlying the validity of our results. Thus, we are confident that our findings are warranted. Second, we only analyzed retroverted hips in which all three signs for retroversion were concomitantly positive (crossover sign, posterior wall sign, ischial spine sign). We did not include patients presenting with only one or two of three criteria. Thus, this is a selected group, even within our own practice. Our conclusions are not generalizable for hips in which only a minor cranial crossover sign with a retroversion index of < 30% is present [39]. We therefore caution that our findings may not be applicable in other practice settings or in less severely affected hips. Third, we used the Merle d’Aubigné-Postel score [6] as the only patient-reported outcome measure. We are aware that there are more appropriate clinical scores available for young patients with FAI [24]. However, most of these new scores have been developed and validated in the past few years and were therefore not available for all our patients from 1997 preoperatively. Unlike other patient-reported outcome measures, the Merle d’Aubigné score was available for all our patients at any followup. We also note that blinding of observers after surgery was not possible because the routine clinical followup evaluation included an assessment of the surgical scar, which inherently would have revealed the type of surgical procedure. Nevertheless, excellent inter- and intraobserver agreement has been reported for the Merle d’Aubigné-Postel score [6] as well as for the assessment of the clinical ROM and the anterior impingement test [25, 26, 55]. Fourth, we do not have standard preoperative MRI for all patients included in the study mostly because patients already presented with an MRI from another institution or the MRIs were no longer retrievable. However, because the demographic and conventional radiographic data of patients with and without MRI were comparable, this should not jeopardize our results. Fifth, the mean followup interval for the rim trimming through surgical dislocation group is smaller. However, this is accounted for statistically by the log-rank test. This test is a time-stratified test that was specifically designed to analyze potential differences of two survivorship curves with different followup intervals. In addition, the shorter followup interval would be in favor of the surgical dislocation group. Inevitably for a study period spanning 15 years, different observers assessed these parameters preoperatively. However, substantial inter- and intraobserver agreement has been reported previously for ROM [12, 26, 55] and the anterior impingement test [25]. Sixth, we have a different rate of patients who either declined routine followup or were lost to followup for the two groups (13% for the anteverting PAO group versus 25% for the rim trimming through surgical hip dislocation group). Because patients lost to followup usually present with similar [5, 18] or worse [14] results, this would strengthen our conclusion that rim trimming for symptomatic acetabular retroversion results in less favorable results compared with anteverting PAO.

Comparing our midterm failure rates with those of others (Table 4), consistently good results have been reported for the anteverting PAO procedure in acetabular retroversion. A direct comparison of the survivorship curves is difficult because various endpoints beyond THA have been used (Table 4). Reports of isolated open or arthroscopic acetabular rim trimming for acetabular retroversion are not available with a followup of 10 years. Most of the reports have a followup of < 5 years [4, 19, 33] or report on “pincer”-type hips without further specification [1, 29, 42]. There is only one study comparing the surgical treatments for acetabular retroversion. In this study, Peters et al. [32] differentiated between the two surgical treatments analyzing two different morphologic types of acetabular retroversion: one with a deficient posterior wall and one with sufficient posterior coverage. The followup after 4 years was comparable regarding the Harris hip score. However, these results have to be considered preliminary because, based on our results, the potential drop in the survivorship curve occurs at Year 6. Our results should be considered as pilot results that offer valuable information to conduct future prospective randomized trials comparing both surgical treatments for acetabular retroversion. We hypothesize that the better outcomes seen with reorientation osteotomy compared with rim trimming occur because the underlying pathology represents a global rotational abnormality of the entire hemipelvis rather than focal overgrowth of the rim [15, 16, 23, 44, 49]. When looking at the surgical treatment of our patient cohort, this is particularly interesting for two reasons. First, none of the patients undergoing anteverting PAO had surgical treatment of the labrum, whereas all patients undergoing surgical hip dislocation underwent labral refixation. Obviously, despite the reported importance of labral refixation [7, 19–21, 37], the patients appear to benefit more from correct acetabular orientation. Second, although the preoperative asphericity of the femoral head-neck junction was comparable in the two groups (Table 2), osteochondroplasty of the femoral head-neck junction was performed less frequently in the anteverting PAO group. Again, this would favor the surgical dislocation group in which asphericity was corrected in all cases.

Table 4 Overview of the literature on treatment of acetabular retroversion
Full size table

In summary, anteverting PAO seems the more appropriate treatment of acetabular retroversion (defined by concomitantly present crossover/ischial spine/posterior wall signs) because symptomatic acetabular retroversion can be treated efficiently without compromising the already small lunate surface area in these hips. Our results are not valid for hips that present with a crossover sign only. Based on the data from others [32], acetabular rim trimming may be a valid option for treatment of hips with a positive crossover sign but concomitantly sufficient posterior coverage.