Perthes disease involves necrosis of the proximal femoral epiphysis and physis. Containment of the necrotic epiphysis within the acetabulum is a common treatment in the early phases of the disease, including: abduction casting, femoral osteotomy, pelvic osteotomy, or shelf acetabuloplasty. After the femoral epiphysis reossifies, some hips have an aspherical and incongruent femoral epiphysis, and several reconstructive–salvage treatments have been reported, including shelf acetabuloplasty, femoral osteotomy, cheilectomy, femoral head reduction, and pelvic osteotomy. Since the shelf acetabuloplasty is used as both a containment procedure and a reconstructive–salvage procedure, we were not clear about the reported outcomes of its usage. The aim of this meta-analysis was to examine published articles referenced in PubMed about Perthes disease treated with shelf acetabuloplasty, with the intent to assess its effectiveness as measured by the Stulberg classification when performed during the early stages of necrosis or fragmentation (i.e., containment procedure) or during reossification or residual stages (i.e., a reconstructive–salvage procedure).
Predictors of outcome of the treatment of Perthes disease include age at onset, extent of epiphyseal necrosis, degree of epiphyseal or hip deformity, and stage of disease at treatment (i.e., containment vs. reconstruction–salvage). Children with Perthes disease who are older than 9 years of age have been shown to have poor outcome compared with younger children [36
]. Children under 6 years of age tend to have good results; however, some hips in young patients may develop severe femoral head deformity [39
]. Fabry et al. [40
], in their systematic review, observed that patients younger than 6 years of age may have poor outcome. Pecquery et al. [18
] found two patients with poor outcome who had shelf acetabuloplasty at less than 6 years of age. Schoenecker et al. [41
] indicated that children younger than 6 years of age may have poor outcome when more than 50 % of the femoral head is involved. In our systematic review, age at the onset of disease was not reported adequately enough to allow a quantitative systematic review. However, poor outcomes (Stulberg IV and V) were reported more commonly in cases in which the shelf acetabuloplasty was performed at an age older than about 10–11 years.
The extent of femoral head involvement in Perthes disease has been shown to be a significant predictor of outcome [42
]. Catterall [43
] described four groups of Perthes disease based on the epiphyseal sequestrum and collapse. Favorable outcomes were expected when less than 50 % of the femoral head was affected by the disease [41
]; total epiphyseal necrosis has the worst prognosis. In our meta-analysis, most cases treated with shelf acetabuloplasty were classified as Catterall III or IV and had substantial necrosis prior to treatment. The degree of epiphyseal or acetabular deformity has been measured by multiple processes including several radiographic indices, including lateral femoral head translation [15
], acetabular coverage [14
], and acetabular angle (Sharp angle) [15
]. We selected the Stulberg classification as the outcome variable.
A qualitative systematic review on the effect of shelf acetabuloplasty in improving hip coverage and decreasing hip symptoms was recently published [44
]. Our meta-analysis revealed that the shelf acetabuloplasty procedure used as containment (performed during Waldenström stages of necrosis or fragmentation) provided a significant, 85 % good Stulberg outcome (classes I, II, or III). However, the shelf acetabuloplasty procedures used as reconstruction–salvage (performed during Waldenström stages of reossification or residual) offered a significant, 69 % good Stulberg outcome (classes I, II, or III). In this meta-analysis, we examined heterogeneity and considered the studies to be homogeneous at a significance level of 2 % (type 1 error tolerance). The random effect meta-analysis is conducted if significant heterogeneity is observed among studies that constitute the meta-analysis. With this approach, it is unlikely that variability between studies will affect our findings. The heterogeneity of studies is one of the most important considerations in meta-analysis. The overall meta-analysis was affected by heterogeneity, which resulted in the use of the random effect meta-analysis of DerSimonian and Laird. The subgroup meta-analysis observing the outcome of shelf acetabuloplasty in early stages of Perthes disease also was affected by heterogeneity, and we utilized random effect meta-analysis in this subgroup as well. Our subgroup meta-analysis that observed outcomes in later stages showed significant homogeneity, and we reported our pool estimate with the fixed effect meta-analysis of Peto and Mantel–Haenszel. The fixed effect meta-analysis is used when there is marginalized variability between studies.
There are a few limitations in this meta-analysis. We used a pool estimate in drawing the inference regarding the outcome of Perthes disease following shelf acetabuloplasty. Like every meta-analysis, heterogeneity remains a limitation of our study. Secondly, because we included only retrospective studies in this meta-analysis, it is possible that our findings may be affected by selection and misclassification biases of the original articles. For example, several studies used the Stulberg classification differently, which has a tendency of introducing selection bias into the original studies and, subsequently, into our meta-analysis. General consensus remains in the application of the Stulberg classification. In terms of reliability, the Stulberg classification has been examined and the level of experience may affect the detection of flattening of the femoral head (i.e., class III vs. class II) [45
]. However, this classification model is widely used and conveys some level of reliability of substantial nature. There is a tendency to classify Stulberg I/II as good outcome, III as fair, and IV and V as poor outcomes. In contrast, I/II/III have also been classified as good, while IV and V have been classified as poor outcome. In our meta-analysis, we applied a classification model that considers I to III as good and IV and V as poor. This classification was applied to all studies equally without any form of misclassification bias. Because the reliability of meta-analysis is contingent on the internal and external validities of the studies selected for the meta-analysis, possibly the lack of confounding adjustment and bias minimization may influence the findings in this meta-analysis. Therefore, to overcome this bias, we used the random effect meta-analysis.
In addition, our meta-analysis is limited because substantial parameters used to assess the outcome of Perthes disease were not available in some of the studies, rendering the pool estimate impossible to compute. Outcomes were appropriate when assessed with a similar instrument used for the baseline.