For the treatment of a dislocated hip after the age of eighteen months, a varus derotation osteotomy or a pelvic osteotomy is commonly performed in addition to open reduction1. The rationale for performing a varus derotation osteotomy is to redirect the femoral head toward the center of the acetabulum with the aim of improving the stability of the concentric reduction, which is the primary stimulus for acetabular remodeling2. Varus derotation osteotomy has been recommended as being technically easier than pelvic osteotomy3,4. Because acetabular remodeling occurs in children eighteen months of age and older5,6, some believe that pelvic osteotomy may be unnecessary in children who are five years of age or less7. Others believe that the capacity for acetabular remodeling declines rapidly after the age of eighteen months and cannot be relied on to reverse the acetabular dysplasia through a stable concentric reduction of the hip alone8. As a result, a pelvic osteotomy is performed at the time of open reduction to correct the residual dysplasia directly8,9.
The choice of osteotomy is controversial10. While some studies have demonstrated improvements in acetabular development following open reduction alone in children older than eighteen months3,5,7,11, it remains unclear which osteotomy is preferable in this age group. Albiñana et al.11 showed that hips that are at risk for long-term dysplasia stop remodeling at an earlier age than those with a better prognosis, which raises the possibility that some hips may be more appropriately treated with pelvic osteotomy than others. In practice, the choice of procedure usually results from the individual surgeon's experience10 rather than from any objective evidence of the superiority of one technique over the other.
The purpose of the present study was to compare acetabular development over a long period of time in patients who were managed for hip dislocation due to developmental dysplasia of the hip after open reduction combined with either varus derotation osteotomy or innominate osteotomy.
We performed a retrospective cohort study at two tertiary care centers of comparable urban teaching hospitals. At one center, all patients with a hip dislocation underwent open reduction combined with varus derotation osteotomy. At the other, all underwent open reduction combined with innominate osteotomy. In each group, one surgeon performed all operations; in one group (open reduction combined with innominate osteotomy), the surgeon was one of the authors (J.H.W.), and in the other group the surgeon was not one of the authors. We studied consecutive patients who were managed during a predefined period from 1990 to 2003. The study was approved by the institutional review boards at both centers. Informed consent was waived as no patient contact was involved in this study. Operating room logs and the individual databases of the two surgeons were used to identify all patients who had been fifteen to forty-eight months of age at the time of open reduction performed simultaneously with either varus derotation osteotomy or innominate osteotomy.
In both groups, open reduction of the hip was performed by the same surgical technique described elsewhere12. Following adductor tenotomy, an anterior approach to the hip was used. The straight head of the rectus femoris tendon was divided, as was the psoas tendon at the pelvic brim. The iliac apophysis was split, and a subperiosteal exposure of the outer table of the ilium was made. A T-shaped capsulotomy was performed, with excision of a lateral flap of capsule, and the hip was reduced. At the end of the procedure, the capsulotomy was closed with use of nonabsorbable sutures. Postoperatively, a hip spica cast was applied for six weeks in both groups, during which time weight-bearing was not permitted.
In the varus derotation osteotomy group, a subperiosteal exposure of the proximal part of the femur was made through a separate lateral incision, with elevation of the vastus lateralis. A medially based wedge of bone was excised in the intertrochanteric region in order to achieve a postoperative neck-shaft angle of 110°, with no additional femoral shortening. Sufficient derotation of the femur was employed to achieve neutral postoperative femoral anteversion of approximately 0° to 5°. Fixation was achieved with use of the Coventry plate and screw system (Figs. 1-A and 1-B).
In the innominate osteotomy group, following subperiosteal exposure of the inner table of the ilium, an innominate osteotomy was performed with use of a Gigli saw. A triangular corticocancellous autograft was harvested from the ipsilateral iliac crest and was inserted in the osteotomy site. Fixation was achieved with use of two heavy threaded Kirschner wires (Figs. 2-A and 2-B). Patients deemed to be candidates for innominate osteotomy who were more than thirty-six months old underwent femoral shortening as well and therefore were not eligible for the present study.
Patients were included if the dislocation was due to developmental dysplasia of the hip and if they had a minimum of four years of follow-up. One investigator at each center (G.S. and R.H.) abstracted medical records to identify diagnosis, sex, side, age at the time of surgery, details of previous treatment, complications (plaster sores or soiling, supracondylar fracture, infection, hardware malposition or failure), removal of hardware, and secondary procedures on the same hip.
Radiographic Assessment
The same radiographic protocol was used for all patients. A supine radiograph of the pelvis, centered on the hips and with both feet in 15° of internal rotation, was made (depending on age) at 60 to 80 kV, at 4 to 40 mA, and with a focus-to-film distance of 150 cm on a digital imaging system (Fujifilm FCR 5000R; Fujifilm, Tokyo, Japan). Preoperative pelvic radiographs made within four weeks before surgery were assessed with regard to the severity of dislocation according to the system of Tönnis13 and with regard to the severity of acetabular dysplasia as measured with the acetabular index14.
Acetabular remodeling was quantified on the basis of the change in acetabular index (the primary outcome) (Fig. 3) over the entire follow-up period (from the first to the latest postoperative radiographs demonstrating an open triradiate cartilage). The acetabular floor thickness15, the distance between the most inferolateral and most inferomedial points of the ilium near the triradiate cartilage, was also measured on these radiographs (Fig. 4). This distance ranges from 8 to 15 mm and has been shown to change over time, with the value continuing to increase until skeletal maturity in hips with a poor outcome11.
We measured hip stability by means of the lateral and superior centering ratios16 (Fig. 5). These ratios quantify femoral head concentricity in the horizontal and vertical planes, respectively17, and are a measure of congruity of the reduction. Normal values are 0.1 to 0.2 for the lateral centering ratio and 0.6 to 0.85 for the superior centering ratio16. Subluxation of the hip was defined as a superior centering ratio of <0.1 on the most recent postoperative radiograph17.
The development of osteonecrosis of the proximal femoral epiphysis after surgery was determined and classified according to the method of Kalamchi and MacEwan18. Six patients with severe (grade-III and IV) osteonecrosis were excluded from the analysis of acetabular development because osteonecrosis of the proximal femoral epiphysis has the potential to affect acetabular remodeling negatively19. Five of these patients, including four with grade-III osteonecrosis and one with grade-IV osteonecrosis, were in the varus derotation osteotomy group, and one patient, who had had grade-III osteonecrosis, was in the innominate osteotomy group (p = 0.17).
To eliminate interobserver variability, all radiographs were graded by the first author (G.S.), who was not involved in any of the surgical procedures. For each patient, all subsequent radiographs that were available in the postoperative course were evaluated. Any patient identifiers were removed, and the radiographs were mixed before grading. Measurements were not made on rotated radiographs showing obliquity of the iliac wings or lumbar pedicles as these radiographs were excluded. Typically, radiographs were available at six weeks; three, six, twelve, eighteen, and twenty-four months; and every year thereafter. On the average, each patient had a set of 6 ± 2 postoperative radiographs. Measurements were made electronically, with use of magnified views and contrast control (Centricity PACS; GE Medical Systems, Chalfont St. Giles, United Kingdom, or Sienet Sky; Siemens Medical Systems, Toronto, Ontario, Canada) when necessary. A small number of patients in both groups only had hard-copy acetate films or a mixture of hard-copy and digital images available for analysis. For hard-copy images, absolute measurements were made with the assumption of a magnification factor of 115% according to the recommendation made by Conn et al.20.
Statistical Methods
Sample-size calculation was based on the primary outcome (the acetabular index), with the aim of detecting a difference of at least 4°, which was considered clinically important3,7; statistical power was set to 80%. We established the reliability of the observer with use of a random set of twenty radiographs. The intraclass correlation coefficient was 0.84 (95% confidence interval, 0.64 to 0.93) for the main outcome, demonstrating "excellent" reliability21. The mean intraobserver difference was 0.65° (95% confidence interval, -1.51° to 2.81°). We performed univariate analyses to compare both groups (the varus derotation osteotomy group and the innominate osteotomy group) for confounders and complications. Categorical variables were analyzed with use of the Mann-Whitney U test or the Fisher exact test as appropriate, whereas continuous variables were analyzed with use of independent-samples t tests.
We used several radiographs to determine the development of the acetabulum over time for each patient and modeled these observations with use of a repeated-measures analysis of variance. Our primary interest was the magnitude of the surgical treatment effect (varus derotation osteotomy or innominate osteotomy) on several radiographic outcomes representing acetabular morphology. Because several such outcomes were available for each patient, together they reflected acetabular remodeling over time. Other variables that we considered to influence acetabular remodeling were sex, age at the time of surgery, the preoperative acetabular index, a history of failed harness treatment and/or closed reduction, the severity of hip dislocation, and the presence of grade-I and II osteonecrosis in the postoperative course. Because the radiographs for each patient were measured at several time points, it was necessary to adjust for correlation between repeated observations within one patient. We used a mixed model to account for this correlation. We decided a priori to adjust all models for age at the time of surgery and the preoperative acetabular index (a measure of the severity of acetabular dysplasia at baseline). Time was treated as a continuous variable, and we included the interaction of time and treatment group (innominate osteotomy or varus derotation osteotomy) in all models. For all models, adding a quadratic term for the covariate "time" improved the fit of the data. The significance level for all analyses was set at the 0.05 level, but we kept covariates with borderline p values of =0.1 in the model. We performed post hoc analyses for all outcomes at six, twenty-four, thirty-six, forty-eight, and sixty months postoperatively. We generated least-square means for both groups for each of these time points. We also calculated differences of least-square means and reported p values and 95% confidence intervals for the differences.
Source of Funding
This study was supported as part of a project grant from the Arthritis Research Campaign, United Kingdom. Funds were used to cover travel expenses incurred as part of the study by one of the authors (G.S.). The funding source did not play a direct role in the investigation.
Thirty-eight patients (forty-seven hips) who underwent open reduction with varus derotation osteotomy between 1998 and 2003 and thirty-three patients (thirty-seven hips) who underwent open reduction with innominate osteotomy between 1990 and 2003 were eligible for inclusion. Six hips in four patients, including five hips in the varus derotation osteotomy cohort and one hip in the innominate osteotomy cohort, demonstrated severe osteonecrosis (p = 0.14); as per our a priori definition, these hips were excluded from further analysis. After the random exclusion of one hip in each of the eleven cases of bilateral involvement, sixty-seven hips in sixty-seven patients were available for further analysis (Fig. 6). The mean duration of follow-up was 6.2 years (range, 4.0 to 9.9 years). A total of 490 postoperative radiographs were analyzed. Patients undergoing varus derotation osteotomy were, on the average, 3.4 months older at the time of the operation (p = 0.05) (Table I).
The mean acetabular index before surgery (and standard deviation) was 40.1 ± 6.0 in the varus derotation osteotomy group and 37.7 ± 5.7 in the innominate osteotomy group (p = 0.1). While the acetabular index immediately improved in children who underwent an innominate osteotomy (mean index at six months, 19.1), it did not do so in children who underwent varus derotation osteotomy (mean index at six months, 33.3). Additional improvement of the acetabular index over time was noted in both groups, whereas the magnitude of improvement was different between the groups (p < 0.0001). Children who underwent varus derotation osteotomy never reached acetabular indices as low as those who underwent innominate osteotomy. At twenty-four months the mean difference in terms of the acetabular index was 12.2 (95% confidence interval, 9.52 to 14.91; p < 0.0001), and at four years it was 9.5 (95% confidence interval, 6.77 to 12.24; p < 0.0001) (Table II). Acetabular remodeling as determined by the change in the acetabular index over time was greatest in the first four years after varus derotation osteotomy. In contrast, after reaching the 15° mark at two years postoperatively, no additional remodeling occurred in the innominate osteotomy group (Fig. 7).
All secondary outcomes were significantly different between the two groups, with better outcomes being found in the innominate osteotomy group (p = 0.03 for acetabular floor thickness, p < 0.0001 for lateral centering ratio, p < 0.0001for superior centering ratio). Acetabular floor thickness increased with time in both groups, but the rate of increase was higher after varus derotation osteotomy than after innominate osteotomy (Fig. 8). The lateral centering ratio remained within the quoted normal range16 for both groups over the time period studied (Fig. 9). The apparently lower values seen after innominate osteotomy arose as a result of the lateral shift of the Perkin line, which increased the denominator in the calculation. The group trends for the superior centering ratio over time differed significantly following varus derotation osteotomy and innominate osteotomy, with that for the varus derotation osteotomy group remaining lower throughout the follow-up period quoted normal range, never achieving the lower limit of the quoted normal range at the end of the follow-up period (Fig. 10).
Other factors besides the type of osteotomy also influenced acetabular remodeling, and these factors were different for each outcome measure studied. A failure of nonoperative treatment of the dislocated hip before the operation influenced the development of the hip in the postoperative course. For most outcomes, except for the "lateral centering ratio" (p = 0.03), this predictor was not significant at the 5% significance level (Table III). Preoperative acetabular index was a significant predictor of acetabular development for all outcome measures used (p = 0.0004 for acetabular index, p = 0.0068 for acetabular floor thickness, p = 0.03 for lateral centering ratio, p = 0.05 for superior centering ratio [Table III]).
Secondary procedures were performed in seven patients in the varus derotation osteotomy group and in four patients in the innominate osteotomy group. In the postoperative period, six procedures were performed for subluxation in the varus derotation osteotomy group, compared with one in the innominate osteotomy group (p = 0.05). Specifically, after varus derotation osteotomy, three hips underwent a shelf procedure at five, six, and eight years postoperatively; two hips underwent revision open reduction (combined with innominate osteotomy one year postoperatively in one hip and combined with revision femoral osteotomy and acetabuloplasty six years after the original operation in the other hip); and one hip underwent revision femoral osteotomy six years postoperatively. One patient had persistent subluxation of the hip two years postoperatively but did not undergo additional reconstruction.
In comparison, one hip in the innominate osteotomy group underwent revision open reduction and varus derotation osteotomy five months postoperatively because of subluxation. One hip in the innominate osteotomy group was shown to have redislocated acutely on the third postoperative day and underwent a closed reduction with the patient under general anesthesia, after which the dislocation did not recur. There were no acute redislocations in the varus derotation osteotomy group. With regard to other secondary procedures, two hips in the innominate osteotomy group underwent open division of intra-articular adhesions through an anterior approach at five and fourteen months postoperatively because of stiffness; both were subsequently shown to have osteonecrosis on plain radiography. In addition, one patient in the varus derotation osteotomy group underwent contralateral epiphyseodesis because of limb-length discrepancy secondary to osteonecrosis eight years postoperatively. Severe (Kalamchi grade-III or IV18) osteonecrosis was seen more frequently in the varus derotation osteotomy group (five of eight cases) than in the innominate osteotomy group (one of four cases).
Operations for the removal of hardware were performed in all hips in the varus derotation osteotomy group and in twenty-six hips in the innominate group (p < 0.0001). All procedures were uneventful. Joint penetration by a pin occurred in one patient in the innominate osteotomy group; this finding was recognized in the immediate postoperative period, and the pin was withdrawn. Other complications included three supracondylar fractures (two in the varus derotation osteotomy group and one in the innominate osteotomy group), five cases of skin excoriation requiring cast changes (three in the varus derotation osteotomy group and two in the innominate osteotomy group), and one superficial wound infection that resolved with oral antibiotics (in the innominate osteotomy group).
The present study evaluated the effect of the type of osteotomy (innominate osteotomy or varus derotation osteotomy) on acetabular development in infants who underwent an open reduction of the hip between the ages of fifteen and forty-eight months. The indications and treatment protocols were identical for both groups, with the exception of the type of osteotomy performed in addition to the open reduction. We found that the type of osteotomy had a significant effect on acetabular development in the postoperative course, which we studied up to the age of eight years.
The acetabular index, a measure of hip dysplasia in children, has been used most widely to grade the success of treatment of developmental dysplasia of the hip7,11,22,23. Normal values have been reported and are widely accepted3,7,24. As expected, in the immediate postoperative period, the acetabular index was significantly lower in the innominate osteotomy group, with a mean difference of 14.3° (95% confidence interval, 11.2° to 17.3°; p < 0.0001), because the defect is directly corrected. However, even after eight years of postoperative follow-up, while the acetabular index in the varus derotation osteotomy group decreased because of acetabular remodeling, it always remained significantly higher than that in the innominate osteotomy group. The remodeling rate was highest in the first four years after varus derotation osteotomy, but hips that were treated with varus derotation osteotomy never reached acetabular indices as low as those in hips that were treated with innominate osteotomy. However, as shown in Figure 7, the group trends showed that the acetabular index improved to 20° five years after open reduction with varus derotation osteotomy. The multivariate analysis showed that a normal acetabular index (<20°) was only found five years after surgery. Harris et al.3 and Lindstrom et al.7 regarded an acetabular index of 20° to 24° to represent a "satisfactory" result. However, the acetabular index decreased to 12° in the innominate osteotomy group at the same time point, a result that the same authors would have regarded as "excellent." It is possible that acetabular remodeling may proceed further in the varus derotation osteotomy group even after eight years of follow-up. Although acetabular remodeling has been reported to continue until the age of eleven years25, we believe that the potential for such remodeling is significantly limited at this age. In a large study of fifty-eight patients, Albiñana et al.11 reported that measurable remodeling, as determined by a significant change in acetabular index, ceases four to six years after reduction. Tasnavites et al.26 found that, following reduction in children at the age of one to two years, the acetabular index improved for approximately three years postoperatively.
Although sequential recording of the acetabular index is commonly employed to assess acetabular development, concerns have been raised in the literature about variability in its measurement. Skaggs et al.27 reported the mean difference of measurements of the acetabular index to range between 2.3° and 5.1°. The 95% confidence interval for intraobserver variability of all measurements was 10.1°, but greater intraobserver variability was noted in hips after a closed reduction, with a 95% confidence interval of combined intraobserver measurements of 10.7°27. Such variability could influence the consistency of repeated measures. In order to ensure consistency across radiographic measurements, we performed reliability studies as a first step. The radiographs were analyzed only after a near-perfect intraobserver reliability was established. The mean difference was 0.65° (95% confidence interval, -1.51° to 2.81°) for the acetabular index. In addition, we employed three other radiographic indices of acetabular development and did not rely on the acetabular index alone when determining the treatment effect.
Cadaver studies also have drawn attention to the potential for error in the measurement of the acetabular index introduced by variable positioning of the pelvis28. With this in mind, it is prudent to include other measures of acetabular development as we did in the present study; for all of these measures, a significant treatment effect could be shown. For example, the floor thickness increased in the varus derotation osteotomy group and remained significantly higher for the entire follow-up period. Although the normal range for floor thickness is not known, Albiñana et al.11 consider a progressive increase in its magnitude to be an important predictor of residual dysplasia.
In both the varus derotation osteotomy group and the innominate osteotomy group, the lateral centering ratio was maintained within the quoted normal range16, which suggests that a concentric reduction was obtained initially. It is difficult to draw conclusions from the comparison of the superior centering ratios in the two groups in the early postoperative period as the values in the varus derotation osteotomy group may be inaccurate because of the intended varus position of the proximal part of the femur, which decreases the ratio. However, at the time of the latest follow-up, by which time the proximal part of the femur had remodeled, the superior centering ratio remained abnormal in the varus derotation osteotomy group, whereas it was maintained within the normal range in the innominate osteotomy group16. A progressive decrease in the superior centering ratio to <0.1 is indicative of the loss of concentric reduction due to migration of the femoral head17. Thus, varus derotation osteotomy appeared to be less effective for maintaining concentric reduction than innominate osteotomy did in the present study. Failure to obtain or maintain a corrective reduction is predictive of persistent acetabular dysplasia11,17, which was reflected in the higher rate of procedures performed for the treatment of residual dysplasia in the varus derotation osteotomy group. Only one case of late subluxation occurred in the innominate osteotomy group, and that case necessitated a femoral derotation osteotomy. This finding implies that, in the majority of cases, the anterior femoral head coverage that is provided by innominate osteotomy may compensate sufficiently for femoral anteversion or that the anteversion remodels to the extent that it is not a cause of late displacement. In contrast, the higher rate of procedures performed for the treatment of dysplasia following varus derotation osteotomy suggests that this osteotomy does not maintain hip stability as well as an innominate osteotomy does.
In order to make predictions about the efficacy of a particular treatment for developmental dysplasia of the hip, such as the type of osteotomy, surrogate outcomes of future degenerative joint disease are commonly employed11,29. Such surrogate outcomes are measures of either acetabular dysplasia or concentricity of the reduction16,30,31. The acetabular index has been shown to be predictive of the Severin classification at skeletal maturity11, which in turn predicts the development of degenerative joint disease29. In their study of hips reduced by open or closed means, Albiñana et al.11 demonstrated that the acetabular floor thickness was strongly associated with the Severin classification. Patients with unfavorable outcomes (Severin grades III and IV) had significantly higher values of acetabular floor thickness than those with favorable outcomes (Severin grades I and II). In the present study, according to all of the criteria that were used to judge acetabular development and hip stability over time, the varus derotation osteotomy group showed less favorable results for as long as eight years postoperatively. As both acetabular index and acetabular floor thickness have been shown to be predictive of future Severin grades, the finding of the present study that innominate osteotomy produces better subsequent acetabular development in this age group may imply that a lower prevalence of degenerative joint disease will occur after innominate osteotomy compared with varus derotation osteotomy. Additional follow-up of these patients is necessary to answer this question definitively.
The concerns about simultaneous open reduction and innominate osteotomy in relation to osteonecrosis32,33 were not confirmed in the present study. Although there was no significant difference in the prevalence of osteonecrosis in the two groups, this may represent a type-II error.
The limitations of the present study should be noted. The study was retrospective in nature, which is associated with the risk of measurement bias. We took steps to mitigate this risk; specifically, radiographic measurements were performed according to a set protocol, reliability studies were done, and all radiographs were reviewed by a single observer to eliminate interobserver reliability effects. Although radiographs were made in a standardized fashion, we encountered some radiographs that could not be graded because of imperfect quality. However, these were few in number. Digital imaging software had been introduced around the time of follow-up, so for some patients it was necessary to use a mixture of digital images and hard-copy radiographs, which potentially leads to additional problems due to inconsistencies in the magnification factors used; this was a particular issue when considering the measurement of the acetabular floor thickness, which requires an absolute measurement rather than an angle or a ratio.
To our knowledge, this is the first comparative study to determine the effect of osteotomy on acetabular remodeling. Innominate osteotomy clearly produced better radiographic indices of hip development for as long as eight years postoperatively, with no increase in the rate of complications. Innominate osteotomy therefore may be the better choice in the context of open reduction of a dislocated hip at walking age. However, additional research is needed to determine the importance of these findings in the long term, especially with regard to the need for secondary procedures and the rates of premature osteoarthritis and femoroacetabular impingement. 
Note: The authors gratefully acknowledge Mr. David Jones, FRCS, for permission to include his former patients in this study.