Background
Some surgeons believe that in the treatment of developmental dysplasia of the hip (DDH), osteonecrosis may be avoided by intentionally delaying a closed or open reduction until the appearance of the ossific nucleus [
1‐
3]. Results of published studies remain inconsistent with some authors advocating a protective effect of the ossific nucleus [
1,
3‐
5] and others demonstrating no effect [
6‐
8].
A previous meta-analysis of six observational studies [
9] concluded that the presence of the ossific nucleus at the time of hip reduction had a protective effect against the development of grade II-IV osteonecrosis according to Bucholz and Ogden [
10] or Kalamchi and MacEwen [
11]. However, this effect was lost when osteonecrosis of any grade was considered. It also showed that in closed reductions an ossified nucleus reduced the risk of osteonecrosis by 60%, whereas no effect was seen in open reductions. Due to the moderate quality of evidence, the meta-analysis identified a need for further research [
9]. With an increase in the number of studies seeking to clarify the effect of the ossific nucleus [
1‐
3,
5‐
8,
12‐
14] we sought to update the meta-analysis.
This study aimed (i) to determine the effect of the presence of the ossific nucleus on the development of osteonecrosis and (ii) to assess whether the type of reduction performed or the grade of osteonecrosis considered relevant would affect the conclusion.
Discussion
Studies about the role of the femoral head ossific nucleus in preventing osteonecrosis following treatment of DDH reported conflicting results, but a 2009 meta-analysis concluded that there was no such effect unless only osteonecrosis grades II-IV were considered [
9]. The meta-analysis, limited due to the small number of studies available at that time, identified a need for an update as further research emerges. We observed a trend in recent years towards delaying treatment of DDH in the absence of an ossified nucleus [
1,
3,
21]. Numerous new publications on this topic encouraged us to update the 2009 meta-analysis in the hope to find robust conclusions.
With 11 studies or 1021 infants, this updated meta-analysis included twice as many studies and three times as many infants as the one published previously [
9]. The primary outcome, development of osteonecrosis of any grade, remained similar with a relative risk of 0.75 (95% confidence interval, 0.46–1.21) in the 2009 study, compared with 0.87 (0.55–1.38) in the present study. We used a random effects model to calculate the primary outcome, as the I-squared value was greater than 50. However, this resulted in a Tau-squared value of 0.27 which shows low heterogeneity between studies. We also performed a fixed effects calculation on the same data set in order to compare results, as the random effect model assumes an average effect of the ossific nucleus on osteonecrosis across all studies. This showed similar results (Additional file
1). For the subgroup analysis we used a fixed effect model as low heterogeneity between studies was identified in the primary outcome meta-analysis, the number of studies was small and we felt it would be impossible to estimate the tau-squared value with any precision. We suggest this is confirmatory evidence that the ossified ossific nucleus does not play a role in the prevention of osteonecrosis – there does not seem to be a causal relationship.
The 2009 study showed a 60% reduction in the relative risk for developing osteonecrosis grade II or worse if the nucleus was ossified at the time of hip reduction (relative risk, 0.43, 95% confidence interval, 0.20–0.90) [
9]. However, this subgroup analysis was based on only four studies [
9]. We were able to update this subgroup analysis by including two additional studies [
4,
8], which were of high quality based on the GRADE criteria. With an increased sample size from 318 to 471 infants, the updated result was no longer statistically significant – there was no effect of the ossified nucleus on the development of osteonecrosis grades II or worse. While the upper bound of the 95% confidence interval for this analysis was 1.08, we suggest that with a point estimate of 0.67 and the relatively high quality of the additional studies, this result is definitely negative.
The 2009 meta-analysis [
9] also showed that the presence of the ossific nucleus reduced the probability of grade I–IV osteonecrosis by 60% (relative risk = 0.41, 95% confidence interval, 0.18–0.91) after closed reduction. This conclusion was based on three available studies or 183 infants. The updated analysis of closed reductions included seven studies or 684 patients showed no longer a protective effect with a relative risk of 0.80 (95% confidence interval, 0.38–1.69).
By contrast, infants treated with closed reductions (253 hips from four studies) [
1,
2,
7,
9] demonstrated a 50% reduction in risk of developing osteonecrosis grade II-IV (relative risk, 0.50; 95% confidence interval, 0.26–0.94). A possible explanation for this finding is that a closed reduction performed on an ossified nucleus represents a lesser insult to a less plastic epiphysis [
2,
3]. It may be that during an open reduction the decreased plasticity of the ossified nucleus is more than offset by the trauma associated with surgery and possible disruption of the capsular blood supply.
We noticed inconsistencies in the reported prevalence of osteonecrosis, which ranged from 6 to 48%. Those studies that reported a prevalence of 6 to 7% [
7,
14,
20] included predominately cases of closed reductions. Further, one study [
7] excluded patients with incomplete notes or imaging, and another study [
20] excluded patients below the age of two months and those over 20 month. Selection bias could possibly account for the low prevalence of osteonecrosis in these studies. Measurement bias is another possible explanation for the wide range in prevalence estimates. Only seven studies (50%) stated the number of outcome assessors [
2,
8,
12‐
14,
19,
20] and only two studies (14%) [
8,
13] evaluated outcomes blinded and included inter-rater reliability studies. Attrition bias was inherent to the study of Pospischill et al. [
13], which excluded 21 patients (21%) due to length of follow-up being less than three years.
All included studies identified the age at reduction as a potential confounder in the relationship between osteonecrosis and an ossific nucleus; however, only ten studies (72%) [
1‐
3,
5‐
8,
12‐
14] provided supportive statistical analyses. Other confounders have been reported: Carney et al. [
1] found an increased risk of osteonecrosis when an adductor tenotomy was omitted as part of a closed reduction (
p = 0.0066). In the study of Pospischill et al. [
13] osteonecrosis was more likely (
p = 0.003) when a concomitant osteotomy was done. Only two studies (14%) performed multivariate analyses [
8,
13].
Our study has limitations. In four studies (29%) the presence of the ossific nucleus was determined solely or in select cases by ultrasound. Because previous research has demonstrated that in 93% of cases an ossific nucleus that is visible on ultrasound is also visible on radiographs [
22] we included these papers in our study. Three studies (22%) [
1,
12,
19] had a follow up of only two years. It could be argued that this might underestimate the prevalence of osteonecrosis; however, two years has been used as an outcome in previous research [
9] and is perhaps acceptable for making reasonable inferences. There was variation in the radiological classifications used for grading the severity of osteonecrosis in different studies. Thirteen studies (93%) used Bucholz and Ogden [
1,
3,
6‐
8,
13,
14,
16] or Kalamchi and MacEwen [
2,
4,
5,
12,
20] which are interchangeable. Madhu et al. [
19] graded osteonecrosis according to Salter et al. [
23] and we thus could include their data only when modeling the effects of the ossific nucleus on osteonecrosis of grades I–IV.
Acknowledgments
All those who contributed to the meta-analysis meet the criteria for authorship and have been listed as authors.