We followed up 76 consecutive hips with symptomatic acetabular dysplasia treated by acetabular shelf augmentation for a mean period of 11 years. Survival analysis using conversion to hip replacement as an end-point was 86% at five years and 46% at ten years. Forty-four hips with slight or no narrowing of the joint space pre-operatively had a survival of 97% at five and 75% at ten years. This was significantly higher (p = 0.0007) than that of the 32 hips with moderate or severe narrowing of the joint-space, which was 76% at five and 22% at ten years. There was no significant relationship between survival and age (p = 0.37) or the pre- and post-operative centre-edge (p = 0.39) and acetabular angles (p = 0.85).

Shelf acetabuloplasty is a reliable, safe procedure offering medium-term symptomatic relief for adults with acetabular dysplasia. The best results were achieved in patients with mild and moderate dysplasia of the hip with little arthritis.

Dysplasia of the hip is a complex developmental process. Untreated acetabular dysplasia is the most common cause of secondary osteoarthritis of the hip and is caused by pathological joint-loading forces.1,2 Options for operative treatment are osteotomy, total hip replacement (THR) and arthrodesis. Joint-preserving acetabular osteotomies include redirectional osteotomies (Salter and Bernese), reshaping osteotomies (Pemberton) and salvage augmentation procedures (shelf, Chiari). The shelf procedure is the oldest acetabular procedure. It was first described by König3 in 1891 and modified and popularised by Gill4 in 1926. Subsequently, Wiberg5 improved the technique and the results of the procedure. With increased interest in redirectional pelvic osteotomies, the role of the shelf procedure now needs to be redefined. Most reports of shelf acetabuloplasty describe its use in children and adolescents2,69 and few have been published on the longer term survival of the procedure in the skeletally mature.1013

Our aim was to investigate the effectiveness of the shelf procedure in adults with symptomatic acetabular dysplasia by assessing the functional and radiological outcome at a minimum follow-up of five years.

Patients and Methods

Between 1986 and 1997, 76 consecutive shelf augmentation procedures were performed by two surgeons (MKDB, PMS) on 67 skeletally mature patients with painful acetabular dysplasia. The pain was aggravated by exercise and daily activities but was often present at rest. Forty procedures were performed on the right and 36 on the left side. There were 53 women and 14 men with a mean age at the time of surgery of 33 years (17 to 60). Six women and three men had bilateral staged procedures. The mean follow-up period was 11 years (6 to 14). The shelf augmentation was combined with a femoral osteotomy in six patients. Seven patients had removal of metal work from previous femoral osteotomies at the same time as the shelf procedure. MRI with arthrography was performed to exclude acetabular labral tears in those with clinical impingement. In 20 patients arthrography was performed pre-operatively and no labral tears were demonstrated in this group.

Standard standing anteroposterior and false-profile pelvic views were taken. Radiological assessment included measurement of the centre-edge (CE) angle of Wiberg and the acetabular angle of Sharp (AA) to determine cover of the femoral head. The severity of the pre-operative osteoarthritis was based primarily on the width of the joint space using the classification of De Mourgues and Patte14 (Table I). Narrowing of the joint space of grade I to grade III was found in 44 hips and of grade IV in 32; none had grade V. It should be noted that the patients were not serially evaluated for the radiological progression of their arthritis. Pain and physical activity, however, were monitored by follow-up and postal questionnaire.

View this table:
Table I.

Classification of osteoarthritis of the hip by De Mourgues and Patte14

Hip congruency and subluxation were assessed.13 In most patients, the hip was subluxed. In only four hips (5%) was the head symmetrically contained in an acetabulum with a short roof in the presence of an intact Shenton’s line. A slight loss of congruency with moderate subluxation was found in 35 hips (46%), and severe loss of congruency with more marked subluxation in 37 (49%). Although congruency was not evaluated by movement studies, radiographs suggested incongruity in the shape of the femoral head and acetabulum.

There was a normal range of movement in 47 hips (63%). The validated Oxford hip score15 (OHS) was used for clinical assessment post-operatively by a postal questionnaire excluding patients who underwent joint replacement.

Survival analysis,16 using conversion to THR as an endpoint, was performed. The patients were divided into two groups, above and below the median value, for each variable of age, pre-operative osteoarthritis and pre-operative and post-operative AA and CE angles. The outcome of each group was compared using the log-rank test (Table II).

View this table:
Table II.

Radiological measurement and the result of log-rank testing

Operative technique.

The operative aim was to stabilise movement of the femoral head by increasing the acetabular cover in the weight-bearing position. The main difference between the technique of the two surgeons was the line of incision. A standard Smith-Peterson incision was used in 48 hips and a bikini incision in 28. The abductors were partially stripped subperiosteally from the outer aspect of the ilium to expose the superior joint capsule. The reflected head of rectus femoris was identified, divided and reflected posteriorly to expose the underlying capsule. Image intensification was usually used for accurate positioning of the graft. Using a drill and an osteotome, a slot was made in the ilium just above the capsular reflection and extended through to the inner table. The exact position of the slot depended on the area of greatest uncovering of the femoral head. A uni- or bicortical bone block was then harvested from the iliac crest and fashioned so that it could be firmly impacted into the slot to achieve stability without internal fixation, using the natural curve of the outer wing of the ilium to cover the femoral head. Further cancellous graft was then packed above the shelf. The reflected head of rectus was not resutured. The abductors were sutured back to the iliac crest.

Initially, the post-operative protocol used skin traction for ten days followed by touch-down weight-bearing for six weeks. However, all recent patients have been mobilised the day after surgery and remain touch-weight-bearing for six weeks. In both protocols partial weight-bearing was allowed at six weeks for a further six weeks followed by full weight-bearing.


The mean pre-operative CE angle from 11° (−20° to +17°) pre-operatively to 50° post-operatively (30° to 70°) and the mean pre-operative AA from 52° (46° to 64°) to 32° (18° to 57°) post-operatively. Six months after operation, pain had improved in 68 (90%) hips. At a mean of five years postoperatively, pain had begun to develop in ten hips (14%). Twenty-two hips (30%) warranted replacement at a mean of 7.5 years (2 to 14) after operation. At ten years after operation, 35 hips (46%) still reported relief from pain.

Before operation the Trendelenburg test was positive in 32 (43%) of hips and in only ten (14%) after operation. The range of movement was not affected by the procedure except in one patient whose graft had been placed too anteriorly with a consequent reduction in flexion. This was subsequently trimmed with an improved range of movement and a good outcome. When hip replacement was necessary, the shelf had united solidly to the ilium in all cases except one.

Survival analysis using conversion to THR as an endpoint was 86% (95% confidence interval (CI) 76 to 95) at five years and 46% (CI 38 to 65) at ten years (Fig. 1). The survival of the 44 hips with narrowing of the joint space of grade I to grade III (Fig. 2), pre-operatively, at five years was 97% (95% CI 93 to 100) and at ten years 75% (95% CI 51 to 100). This was statistically significantly higher (p = 0.0007) than the survival of the 32 hips with narrowing of grade 4, which at five years was 76% (95% CI 55 to 89) and at ten years 22% (95% CI 5 to 38) (Fig. 3). Thirty-nine hips with congruent (four) or only mildly incongruent hips (39) showed a survival of 94% (95% CI 82 to 100) at five years and of 65% at ten years (95% CI 59 to 66). This was statistically significantly higher (p = 0.0387) than that of the 37 patients with moderate or severely incongruent hips. Their survival at five years was 89% (95% CI 75 to 100) and at ten years 35% (95% CI 10 to 61) (Fig. 4). There was no significant relationship between survival and age (p = 0.37; Fig. 5), pre-operative CE angle (p = 0.39; Fig. 6), postoperative CE angle (p = 0.27); pre-operative AA (p = 0.85; Fig. 7) and post-operative AA (p = 0.87).

Fig. 1

The survival curve for shelf arthroplasty.

Fig. 2a, Fig. 2b

Radiographs showing a) pre-operative acetabular dysplasia with grade 3 osteoarthritis and b) the appearance 10 years after shelf arthroplasty.

Fig. 3

The relationship between the survival of the shelf arthroplasty and osteoarthritis (OA).

Fig. 4

The relationship between the survival of the shelf arthroplasty and hip congruency (bars indicate 95% confidence intervals).

Fig. 5

The relationship between the survival of the shelf arthroplasty and age (bars indicate 95% confidence intervals).

Fig. 6

The relationship between the survival of the shelf arthroplasty and pre-operative centre-edge angle (CEA) (bars indicate 95% confidence intervals).

Fig. 7

The relationship between the survival of shelf arthroplasty and pre-operative acetabular angle (AA) (bars indicate 95% confidence intervals).

Excluding the group who had undergone hip replacement, responses to a postal questionnaire were available for 43 hips (80%). Of these, the mean OHS was 35 (hip score maximum, 48).

Complications (Table III).

View this table:
Table III.

Complications of shelf arthroplasty

Ten patients had meralgia paraesthetica. The Smith-Peterson incision was used in eight and the bikini incision in two. Only five patients at the last follow-up had residual dysaesthesia and none regarded it as significant. There was no other neurological injury. Four patients had nonunion of a broken graft. One patient aged 57 years and previously asymptomatic, fell heavily five years after the operation and fractured the graft. One patient had Munchausen’s syndrome diagnosed by a psychiatrist. The four patients who had nonunion of the graft had revision and screw fixation. Two had a good result. One suffered breakage of screws and required a THR. The fourth had a further nonunion and no further surgery since the symptoms did not warrant it. Two of the six patients who had undergone a femoral osteotomy had a painful bursa, which resolved when the plate was removed.


Hip dysplasia in untreated patients generally progresses to disabling arthritis. Stulberg and Harris17 recognised the strong association between acetabular dysplasia and osteoarthritis. They found that of 130 selected patients with osteoarthritis, 48% had acetabular dysplasia. Wiberg’s18 finding of an incidence of acetabular dysplasia of 25% in those with osteoarthritis still remains widely accepted. Surprisingly, Croft et al19 suggested that acetabular dysplasia was unlikely to be an important cause of osteoarthritis in men since there appeared to be little relationship between narrowing of the joint-space and the CE angle. In addition, there is no available information about the rate of progression of osteoarthritis in acetabular dysplasia.

Our encouraging results in adults with mild arthritis are comparable with those of other series, although there are few reports in the English literature. Rosset et al12 suggested a survival probability of 50% after 25 years in a group of 68 adults of mean age of 33 years, undergoing shelf arthroplasty. A total of 77% of the patients had arthritis of grade 3 to grade 4 and 23% of grade 1 to grade 2. Migaud et al13 showed that the survival rate of 56 painful dysplastic hips (mean age of patients 32 years) treated by shelf arthroplasty was 58% at 15 years and 37% at 20 years. Failures were analysed in terms of a number of variables. There appears to be general agreement about the relationship between pre-operative narrowing of the joint-space and survival. This was confirmed by our study. We are now guarded in recommending shelf arthroplasty for patients with marked arthritis. Our study showed no relationship between survival and the age of treatment or the initial CE angle. This supports the findings of Rosset et al12 but differs from those of Migaud et al.13 There is controversy regarding the significance of age and the severity of dysplasia (CE angle).1214,2023

Dysplasia of the hip unfavourably alters the normal hip biomechanics.24 The development of premature arthritis in patients within acetabular dysplasia is in part mechanical because of increased contact stress. In addition, there are shearing forces between the femoral head and the shallow acetabulum. This leads to stretching of the capsule with mechanical pain which is worse on walking and exercise rather than at rest. In the shelf procedure, stabilisation of the femoral head helps to prevent the shearing forces which accompany weight-bearing and allows a more even distribution of articular pressure. This may protect against the accelerated arthritis which accompanies subluxation and instability. Similar aims are achieved with the Chiari osteotomy, but the shelf procedure is less technically demanding and has fewer complications.25 However, it is acknowledged that in isolation it does not centralise the hip and like the Chiari osteotomy remains a salvage procedure. It relies on fibrocartilaginous metaplasia of the weight-bearing superior capsule. Fibrocartilage is less durable than hyaline cartilage.

Placing the graft at the right level should contribute to stability of the hip. When this occurs graft hypertrophy develops. It has been shown that if the graft is placed too high it may resorb with time and if placed too low may cause intraarticular damage.6 If the graft is not of adequate quality weight-bearing may allow it to fracture. Four of our patients were in this category, one following a fall.

It has been recognised that acetabular labral tears may influence the result of surgery for dysplasia.26 Asymmetrical chronic shear stresses of the rim contribute to labral tearing. There is a strikingly increased incidence of the acetabular rim syndrome in association with dysplasia.27 Not every patient had pre-operative MRI or arthrography. Of the 23 patients regarded as failures, 18 did not have this investigation. Several patients may have had a shelf arthroplasty above a labral tear which may have influenced the result adversely.

Our retrospective study has inevitable weaknesses. The OHS was not recorded pre-operatively and there was no control group or comparison with other osteotomies.

Nevertheless, we have confirmed that shelf arthroplasty is a safe, relatively simple option for patients with mild and moderate acetabular dysplasia and minimal secondary arthritis. When hip replacement becomes necessary, a previous shelf arthroplasty can improve acetabular bone stock and facilitate replacement.


  • No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article.

  • Received August 2, 2004.
  • Accepted March 30, 2005.


View Abstract