Mechanical failure of the hip joint occurs when otherwise normal articular cartilage and bone are subjected over time to excessive unit loads 1-3 . It has been postulated that timely elimination of both abnormal joint-loading and excessive contact stresses may prevent the onset of osteoarthrosis. Even in a joint that already has some degree of mechanically based osteoarthrosis, reduction of the contact stress may reduce the severity of the osteoarthrosis 2-4 .
Many types of pelvic osteotomies were designed to restore more normal mechanics in dysplastic hips 5-19 , but application of some of these procedures to adolescents or adults is difficult. The single innominate (Salter) osteotomy is limited in the degree of correction allowed because of loss of elasticity of the symphysis pubis 2,6 . The ischial cut of the Steel triple osteotomy is quite far from the joint and may result in ischial deformity and nonunion when large angular corrections are attempted.
Periacetabular osteotomies include those described by Eppright 12 , Ganz et al. 13 , Ninomiya and Tagawa 14,15 , and Wagner 16 . The Bernese periacetabular osteotomy described by Ganz et al. involves a series of straight cuts and a controlled fracture to separate the acetabulum from the surrounding pelvis 13 . The advantages of this approach are that the acetabular fragment is large enough to minimize the risk of avascular necrosis and the osteotomy site is stable, obviating the need for a hip spica cast. The disadvantages of the Bernese osteotomy are that not all of the osseous cuts are performed under direct vision and mobilization of the acetabulum tends to cause anterior displacement of the hip-joint center because of asphericity of the osteotomy surfaces. Additionally, there is a substantial so-called learning curve 11 . The spherical acetabular osteotomies described by Wagner and by Ninomiya and Tagawa have the advantages of excellent congruity at the site of the osteotomy, rapid healing, and intrinsic stability with use of only minimal internal fixation. However, these osteotomies are usually performed through the Smith-Petersen anterior approach so that the capsule of the joint can be kept intact to preserve an adequate blood supply to the isolated acetabular fragment. These osteotomies also leave the acetabular teardrop in its original position and therefore are intra-articular at their distal-medial extent 2,14,15 . Spherical acetabular osteotomies also have a steep learning curve.
In this paper, we report a modification of the spherical acetabular osteotomy with use of a modified Ollier transtrochanteric approach. We hypothesized that, compared with other periacetabular osteotomies, this approach would be more readily learned and result in fewer instances of ischemic necrosis of the acetabular fragment and intra-articular penetration.
Between March 1991 and June 1999, thirty-seven patients (thirty-nine hips) were treated with periacetabular osteotomy through a modified Ollier transtrochanteric approach. One patient was lost to follow-up. The remaining thirty-six patients (thirty-eight hips) were included in the study. The study was approved by our Institutional Review Board, and all patients gave informed consent to participate. All except one patient were female, and their ages ranged from fourteen to fifty-three years (mean and standard deviation, 29.42 ± 9.1 years). Twenty right hips and eighteen left hips were involved.
Clinical evaluation was performed preoperatively with a specific focus on pain, previous management, limping, numbness, daily activity, and symptoms of acetabular rim syndrome (a sharp pain in the groin with the hip in passive flexion, adduction, and internal rotation 20 ). Modified Harris hip scores were recorded 21,22 . Radiographic evaluation included an anteroposterior radiograph of the pelvis and a false-profile radiograph of the hip 23 . Radiographic measurements included the anterior center-edge angle, the lateral center-edge angle, the acetabular index of the weight-bearing zone (acetabular index angle), the acetabular head index (lateral coverage of the acetabulum), lateral subluxation (measured from the lateral side of the teardrop to the medial edge of the femoral head), and the status of the Shenton line 23-26 .
The degree of acetabular dysplasia was graded according to the Severin classification 27 . The severity of the osteoarthrosis was staged radiographically according to the criteria of T�nnis 24 .
Indications for the osteotomy were symptoms secondary to acetabular dysplasia that had not responded to at least six months of conservative treatment. The procedure was contraindicated for young patients in whom the acetabular triradiate cartilage was still open with growth potential of more than one year, for patients who had severe acetabular dysplasia of Severin Class VI, and for patients who had osteoarthrosis of T�nnis Grade 2 or 3 and were older than fifty years of age. All operations were performed by the same surgeon.
Operative Technique
The modified Ollier approach has been used for complicated hip surgery, including total hip arthroplasty, at Jikei University, Tokyo, for more than fifty years 28 . Good exposure of the hip joint and easy performance of the operation are the hallmarks of this approach, which is also recommended for surgical treatment of complex acetabular fractures 29,30 . We began to use the modified Ollier approach for periacetabular osteotomy in 1991. After induction of general anesthesia, the patient is placed in a lateral decubitus position and the involved hip and leg are prepared in a sterile fashion. A u-shaped skin incision (modified Ollier incision) is made beginning at the anterior superior iliac spine, curving distally to about 3 cm distal to the greater trochanter, and ending at the posterior superior iliac spine ( Fig. 1-A ). After soft-tissue dissection, the anterior limb of the approach is entered between the plane of the gluteus medius and the tensor fascia femoris. The posterior limb is entered by blunt splitting of the fibers of the gluteus maximus at the level of the posterior border of the gluteus medius to expose the short external rotator muscles of the hip joint. After marking of the greater trochanter and the proximal part of the femoral shaft for later reattachment, an osteotomy of the greater trochanter is performed with care to avoid injury to the femoral neck. The greater trochanter is then turned proximally along with the gluteus medius and minimus and part of the short external rotators. The tensor fasciae latae are protected and not cut. This circumferential dissection is continued to expose about 2 cm of the innominate bone.
The periacetabular osteotomy is performed with a curved osteotome, designed to approximately correspond to the circumferential curvature of the acetabulum ( Fig. 1-B ). The osteotomy begins anterosuperiorly and superolaterally. The line of the osteotomy lies about 1.5 cm proximal to the acetabular edge and follows the attachment of the capsule to the pelvis. Posteriorly, the space between the greater sciatic notch and the joint space is about two fingerbreadths. The osteotomy line bisects this space. The anteroinferior portion of the osteotomy connects the previous line and lies on the protuberance of iliopubic bone. The exposure of the protuberance of the iliopubic bone may be facilitated by dissection between the sartorius and the tensor fasciae latae and by cutting the iliopsoas tendon. After retraction of the quadratus femoris, the curved osteotome is extended anteriorly to cut the posteroinferior part of the acetabulum. The most inferior part of the acetabulum is then fractured or cut directly without penetrating the hip joint.
The acetabular fragment is then displaced anteriorly, inferiorly, and laterally under direct vision to gain adequate coverage of the femoral head with the posterior-inferior portion of the rotated fragment contacting the pelvic wall ( Fig. 1-C ). Two or three Kirschner wires and/or cortical screws are then inserted from the anterior inferior iliac spine and directed proximally toward the sacroiliac joint to transfix the rotated acetabular fragment to the ilium. The hip joint is flexed and abducted to make sure that there is no excessive anterior or lateral correction leading to acetabulofemoral impingement. A corticocancellous bone graft is taken from the external iliac wing and impacted into the space created between the rotated fragment and the ilium. Allograft bone can also be used. An intraoperative radiograph is made to confirm that displacement of the fragment, coverage of the femoral head, and congruence of the hip joint are appropriate.
If there has been relative overgrowth of the greater trochanter due to growth arrest of the capital femoral physis so that its tip is near the joint line, a distal transfer of the greater trochanter is indicated. The recipient site on the lateral surface of the proximal part of the femoral shaft is prepared with a bone saw, creating a flattened surface about 2 cm distal to the original osteotomy level (compared with the contralateral side). With the limb in abduction and the previous marking as a guideline, the greater trochanter is displaced distally and laterally. Two 4.5-mm cancellous screws are inserted for fixation. If there is no need for distal transfer, the greater trochanter is reattached to the original site with two cancellous screws. The Kirschner wires are imbedded subcutaneously with the wound closed over suction drains.
Postoperative rehabilitation includes bed rest for three, four, or five days. Quadriceps and hip-abductor strengthening exercises as well as range-of-motion exercises of the lower limb are undertaken gradually. The patient is discharged one week postoperatively with wheelchair activity allowed. Indomethacin, 75 mg daily, is used to prevent heterotopic ossification for the first six weeks after surgery. The Kirschner wires are removed eight weeks after the operation. Crutch-walking with partial weight-bearing is started one week postoperatively, and full weight-bearing is allowed twelve to sixteen weeks after surgery.
Clinical Data
Patients were assessed preoperatively; at one, two, four, six, and twelve months postoperatively, and then annually. At the latest follow-up evaluation, modified Harris hip scores 21,22 were determined and an anteroposterior radiograph of the pelvis and a false-profile radiograph of the hip were made ( Figs. 2-A , 2-B , and 2-C ). With use of an outcome system combining the modified Harris scores and an assessment of the radiographic severity of osteoarthrosis, we graded the result as excellent when the Harris score was 90 to 100 points and the radiographic grade had remained constant or had improved, good when the Harris score was 80 to 89 points and the radiographic grade had remained constant or had improved, fair when the Harris score was 70 to 79 points or when the radiographic grade had deteriorated and the Harris score was 70 points, and poor when the Harris score was <70 points with or without deterioration of the radiographic grade. An excellent or good result was considered a satisfactory outcome, and a fair or poor result was considered an unsatisfactory outcome.
Data, which were entered and analyzed with the SAS program (SAS Institute, Cary, North Carolina), were represented as the mean, standard deviation, and median for continuous variables and as a percentage for discrete variables. Three major analyses were performed. In the first, preoperative measurements were compared with postoperative measurements, with use of the Wilcoxon signed-rank test and the p value set at 0.05 for each test before the analysis. The second analysis was done to test any possible significant risk factors associated with an unsatisfactory outcome. The nonparametric Wilcoxon rank-sum test was used to assess the difference between the rates of satisfactory and unsatisfactory outcomes for each continuous variable, and the Fisher exact test was used for each discrete variable. The p value was set at 0.01 for each test before the analysis. The third analysis explored any differences between the results for the first nineteen hips treated in the series (Group I) and those for the nineteen hips treated later in the series (Group II) to assess the "learning curve" with the operation. The Wilcoxon rank-sum test was used to compare differences between the two groups for each continuous variable, and the Fisher exact test was used for each discrete variable. The p value was set at 0.001 for each test before the analysis.
The pertinent data for each patient are presented in Tables I and II . The modified Harris hip scoring system was separated into three main categories (pain, function, and range of motion) for further analysis. The function category included limp, use of walking support, distance that could be walked, stair-climbing, sitting, putting on socks, tying shoes, driving an automobile, and absence of deformity. At a mean of five years and six months (range, two to ten years) after the operation, pain had decreased in thirty-four hips, had not changed in three, and had worsened in one. The mean pain score improved from 20.3 ± 9.7 points preoperatively to 38.6 ± 8.5 points postoperatively (p = 0.0001) (see Appendix).
Functionally, limping for more than four months postoperatively was noted after eleven of the hip procedures. The mean functional score improved from 31.8 ± 8.0 points preoperatively to 41.9 ± 6.3 points postoperatively (p = 0.0001). All patients had a satisfactory range of motion. On the average, flexion of the hip improved from 92.0° ± 11.7° preoperatively to 100.1° ± 14.0° postoperatively, abduction improved from 22.5° ± 9.8° to 32.2° ± 9.4°, adduction improved from 12.9° ± 7.5° to 24.9° ± 12.5°, external rotation improved from 12.4° ± 5.0° to 18.0° ± 6.2°, and internal rotation improved from 7.9° ± 4.7° to 13.4° ± 5.9°. The mean score for range of motion improved from 7.1 ± 1.0 points to 7.6 ± 0.8 points (p = 0.0013). The mean total Harris hip score improved from 59.1 ± 15.8 points to 88.0 ± 14.3 points (p = 0.0001).
The mean anterior center-edge angle improved from 22.0° ± 12.9° preoperatively (excluding nine patients who did not have suitable false-profile radiographs for preoperative measurement of the anterior center-edge angle) to 36.1° ± 12.3° postoperatively (p = 0.0001), the mean lateral center-edge angle improved from -2.7° ± 14.4° to 26.6° ± 14.1° (p = 0.0001), the mean acetabular index angle improved from 23.4° ± 6.6° to 12.7° ± 4.6° (p = 0.0001), and the mean acetabular head index improved from 48.2% ± 12.7% to 73.1% ± 16.0% (p = 0.0001). The degree of osteoarthrosis decreased in eleven hips, remained unchanged in twenty-four, and increased in three.
Overall, twenty-five patients (twenty-six hips) had an excellent result, five patients (six hips) had a good result, two patients (two hips) had a fair result, and four patients (four hips) had a poor result. Thirty patients (thirty-two hips; 84%) had a satisfactory result, and six patients (six hips) (Cases 5, 7, 20, 27, 28, and 29) had an unsatisfactory result. Six patients (Cases 2, 9, 15, 20, 27, and 29) had poliomyelitis. The index operation was performed in the nonparalytic limb in four of these patients (Cases 2, 15, 27, and 29) and on the paralytic limb in two (Cases 9 and 20). Of these six patients, three (Cases 2, 9, and 15) had a satisfactory result and three (Cases 20, 27, and 29) had an unsatisfactory result. If the six patients with neuromuscular hip dysplasia due to poliomyelitis are excluded, twenty-seven patients (twenty-nine hips; 91%) had a satisfactory result and three (three hips) (Cases 5, 7, and 28) had an unsatisfactory result.
The possible risk factors that were analyzed to determine whether they were associated with an unsatisfactory outcome are shown in Table III . Age; operative time; blood loss; duration of follow-up; preoperative pain, function, and range of motion; and preoperative radiographic parameters were analyzed, but only preoperative function had a significant association with a poor outcome (p = 0.00191). When the earlier cases (Group I) and the later cases (Group II) were compared regarding these factors, only operative time differed (214 ± 37 compared with 172 ± 25 minutes) (p = 0.00045).
Four patients (Cases 6, 10, 14, and 25) had an acetabular rim syndrome preoperatively and needed a partial arthrotomy to debride the labrum. Each of these four patients had a satisfactory result. Five patients (Cases 11, 13, 18, 19, and 33) had a distal transfer of the greater trochanter because of coxa breva, and all had a satisfactory result ( Figs. 3-A and 3-B ). Two of these patients (Cases 28 and 31) had had prior surgery on the involved hip at the age of eight and seven years, respectively; two patients (Cases 18 and 32) had had prior closed reduction of the involved hip at the age of one year; and two patients (Cases 12 and 16) had had a different previous operation on the contralateral hip.
Complications included prolonged limping (eleven hips); numbness in the distribution of the lateral femoral cutaneous nerve (four); osteonecrosis of the rotated acetabular fragment (two); and acetabulofemoral impingement, heterotopic ossification, and a defect on the rotated ilium (one hip each). Eleven patients (Cases 2, 5, 7, 9, 15, 17, 20, 27, 28, 29, and 36) limped for more than four months postoperatively. Preoperatively, three of these patients (Cases 2, 17, and 28) had had a mild limp and the others had had a severe limp. Six of the patients who limped had had poliomyelitis. One of these six patients (Case 9) no longer limped two years postoperatively, three (Cases 2, 15, and 29) continued to have a mild limp at the time of the latest follow-up, and two (Cases 20 and 27) continued to have a severe limp. Of the remaining five patients, one (Case 36) stopped limping six months postoperatively, one (Case 28) stopped at one year, one (Case 17) stopped at two years, one (Case 5) had a persistent limp due to worsening of the osteoarthrosis and mild weakness of the hip abductors, and one (Case 7) had osteonecrosis of the acetabular fragment and was treated with a total hip arthroplasty two years and three months after the index operation. The two patients who had osteonecrosis of the rotated acetabular fragment had used steroids (Case 7) or had a history of alcoholism (Case 27). There was no case of deep-vein thrombosis or of nonunion of the osteotomy site in our patients.
Periacetabular osteotomy is an effective procedure for the correction of dysplasia, and if it is performed early in the course of osteoarthrosis it can restore function, provide pain relief, and halt or retard the osteoarthrotic process. Each type of periacetabular osteotomy has advantages and disadvantages 31-35 . Trousdale et al. 25 reported that thirty-two of thirty-three patients who had had Grade-1 or 2 osteoarthrosis had an excellent or good result after the Bernese periacetabular osteotomy described by Ganz et al. 13 . The mean lateral center-edge angle improved from 3° preoperatively to 31° postoperatively. However flexion of the acetabulum tended to cause anterior displacement of the hip joint because of asphericity of the osteotomy surface 2 . With the Ganz procedure, assessment of anterior correction can be difficult and anterior overcorrection can occur 32 . Also, the Bernese osteotomy has a major learning curve. Davey and Santore 11 described Santore's preparation for his first periacetabular osteotomy, which included two on-site visits with an experienced surgeon and the performance of nineteen cadaver procedures during a one-year period.
We believe that good rotation and anatomical and biomechanically sound correction can be achieved by a spherical periacetabular osteotomy as described by Wagner 16 , Eppright 12 , and Ninomiya and Tagawa 14,15 . The disadvantages of these types of osteotomies, however, include the possibility of penetration of the joint and devascularization of the acetabular fragment 2 .
When a greater trochanteric osteotomy through a modified Ollier approach is used, the blood supply of the gluteus medius is not disturbed and the viability of the isolated acetabular fragment is better preserved, even when partial capsulotomy is performed 36 . Union of the greater trochanter was not a problem in our series, and distal transfer of the greater trochanter could be performed simultaneously in patients with coxa breva. Also, we adopted a modified (wider) osteotomy through the pubic ramus, which avoided joint penetration and minimized the risk of ischemic necrosis of the acetabular fragment.
After a mean duration of follow-up of five years and six months, the range of hip motion had improved significantly (p = 0.0013). This finding suggested that some patients had painful limitation of the hip joint, instead of contracture, before the surgery and, furthermore, that our approach did not impair hip joint motion. Eleven patients, including the six who had neuromuscular hip dysplasia due to poliomyelitis, limped for more than four months postoperatively. Of the five patients who did not have poliomyelitis, three eventually stopped limping. Of the other two, one had a persistent limp because of worsening of the osteoarthrosis and the other had osteonecrosis of the acetabular fragment. While substantial abductor weakness has been noted after use of a classic iliofemoral exposure or longitudinal incision 15,37 , Murphy and Millis 37 reported recovery of abductor function to preoperative levels by three months postoperatively when an abductor-sparing approach had been used for periacetabular osteotomy. Ninomiya and Tagawa 15 cautioned that obvious weakness of the abductor muscles might be a contraindication to a periacetabular osteotomy because of the risk of a severe postoperative limp. We think that the prolonged limping in our series might have been caused by preoperative weakness of the hip abductors or temporary weakness of the abduction mechanism during healing of the iliotibial band and the site of the greater trochanteric osteotomy. Early operative intervention before weakness of the hip abductors develops and aggressive strengthening exercise of the hip abductors are suggested to prevent prolonged postoperative limping.
We analyzed many factors to determine whether any were associated with an unsatisfactory outcome, but only poor preoperative function was significantly related (p = 0.00191). This finding indicated that painful dysplastic hips should be treated before function is severely impaired. Five other possibly unfavorable factors, including severe dysplasia with a pseudoacetabulum, severe (Grade-3) preoperative osteoarthrosis, previous surgery on the dysplastic hip, musculoskeletal disease such as poliomyelitis, and factors predisposing to the development of osteonecrosis, were not found to be significantly related to an unsatisfactory outcome, although the numbers of patients in the subgroups were small. We believe that a single risk factor is not a contraindication to periacetabular osteotomy but the procedure is probably not indicated for patients with more than two unfavorable factors. Also, these five factors are associated with a poor preoperative functional score. In our series, two patients (Cases 20 and 29) with poliomyelitis had an unsatisfactory outcome because they continued to fall to the ground occasionally after the osteotomy as a result of the paralytic disorder. Nonetheless, they were satisfied with the surgical result because it seemed that there was no better option to correct the painful dysplastic hip.
With our approach, the periacetabular osteotomy can be performed under direct vision and the acetabular fragment can be rotated to any position desired. The hip joint can be flexed and abducted intraoperatively to check alignment and prevent anterior or lateral overcorrection. Acetabulofemoral impingement occurred in only one patient (Case 20), who had an anterior center-edge angle of 56° postoperatively. The acetabular fragment was intentionally rotated excessively to stabilize a paralytic hip in this patient. Crockarell et al. 38 reported that the mean anterior center-edge angle was 32.8° (range, 17.7° to 53.6°) in a study of nine female and thirty male cadavera. They suggested that the threshold of abnormality of the anterior center-edge angle might be slightly lower than previously thought. Our data showed that the anterior center-edge angle might be as large as 54° in a patient without acetabulofemoral impingement.
When the patients treated earlier in the series were compared with those treated later, operative time was the only variable that was found to differ (p = 0.00045). There were no significant differences between the preoperative modified Harris hip scores or between the radiographic results.
Schramm et al. 33 reported that, at a mean of seventeen years after Wagner spherical acetabular osteotomies in thirty-eight hips, 54% of their patients had a good functional result and progression of arthritis was prevented in 63%. Siebenrock et al. 35 reported that, at a mean of 11.3 years after seventy-one symptomatic dysplastic hips were treated with the Bernese periacetabular osteotomy, fifty-eight (82%) had preservation of the hip joint and 73% had a good or excellent result. Major complications included an intra-articular cut in two patients, excessive lateralization in one, secondary loss of correction in two, and subluxation of the femoral head in three.
Trumble et al. 31 reported that, at a mean of 4.3 years after 123 periacetabular osteotomies in 115 patients, 83% of the hips were rated as good or excellent clinically. The majority of complications, which included arterial thrombosis in three patients and laceration of the femoral vein in one, occurred when the osteotomy had been performed through the ilioinguinal approach. Crockarell et al. 32 reported the results of periacetabular osteotomy in twenty-one hips (nineteen patients) followed for more than two years. Complications included two peroneal palsies, three ischial fractures, three asymptomatic pubic nonunions, and five cases of asymptomatic heterotopic ossification. In the current series, thirty patients (thirty-two hips; 84%) had a satisfactory result. Complications included prolonged limping (eleven hips); numbness in the distribution of the lateral femoral cutaneous nerve (four); osteonecrosis of the rotated acetabular fragment (two); and acetabulofemoral impingement, heterotopic ossification, and a defect on the rotated ilium (one hip each).
We concluded that periacetabular spherical osteotomy through a modified Ollier approach prevents joint penetration, usually preserves vascularity to the acetabular fragment, allows partial capsulotomy when necessary, and allows distal transfer of the greater trochanter when necessary. The procedure is easier to perform than most acetabuloplasties, and the clinical results appear to be equal to or better than those achieved in other reported series 31-35 .
A table comparing many preoperative and postoperative variables in the patients in this study is available with the electronic versions of this article, on our web site at www.jbjs.org (go to the article citation and click on "Supplementary Material") and on our quarterly CD-ROM (call our subscription department, at 781-449-9780, to order the CD-ROM).
Note: The authors thank Professor Dennis R. Wenger for the valuable suggestions regarding the manuscript and the deceased Professor Katayama Ry�suke and Professor Murota Kagehisa for their introduction of the procedure to the armamentarium of orthopaedic science.