Introduction

Since osteonecrosis of the femoral head (ONFH) usually affects young adults, various treatments have been developed to prevent or delay the progress of the disease [1]. Core decompression (CD) has been the most common treatment for that purpose over the past 30 years, because it is simple, minimally invasive, easy to perform, and does not require any special equipment [2]. To enhance bone regeneration, bone graft (BG) and bone marrow mononuclear cells (BMMC) have been combined to traditional CD. However, there has been debate over the efficacies of these modalities because inconsistent outcomes with varying success rates have been reported [3,4,5].

To date, four meta-analyses have been performed to determine the effectiveness of CD and concluded that CD was effective when it was done in the earliest stage ONFH [6,7,8,9]. However, all of these meta-analyses had weak statistical strength because they included non-randomized clinical trials or observational studies and did not evaluate the efficacies of adjunctive CD modalities. Moreover, these analyses pertained to pairwise comparisons between two treatments and did not provide comprehensive and overall evidence about the effectiveness of various CD modalities and non-operative treatment.

The natural course of ONFH has been well studied [10]. The size of necrotic portion is the major determinant of further collapse of the necrotic femoral head. The collapse does not occur in small lesions even without any intervention. Thus, small pre-collapse lesions do not progress to further stage and remain in early stage without collapse of necrotic femoral head. It should be argued that previous studies, which reported preventive effect of CD, might have included small early lesions and resultant selection bias.

The network meta-analysis (NMA) or multiple-treatments meta-analysis is a recently developed methodology that allows the evaluation of multiple treatment modalities. It compares direct and indirect evidences on relative effectiveness of numerous treatments and provides a relative ranking of all treatments in terms of effectiveness with a reliable power [11].

When there are multiple treatments for a certain disease, and only a few pair-wise comparisons between two treatments are available, the NMA allows the estimation of efficacies of the multiple treatments, regardless of whether there have been direct (head-to-head) comparisons including all treatments. As far as the effectiveness of treatments for ONFH is concerned, the NMA application allows direct and indirect comparisons among various treatments.

In an attempt to compare the efficacy of various CD modalities and non-operative treatment in patients with ONFH, we conducted a NMA. The efficacy was investigated by comparing the rates of (1) conversion total hip arthroplasty (THA), (2) radiographic progression, and (3) pain relief.

Materials and methods

Search strategy and selection of studies

This systematic review and network meta-analysis was conducted according to the updated preferred reporting items for systematic review and meta-analysis (Electronic supplementary material (ESM) 1) [12], and guidelines for network meta-analysis [11]. In November 2016, a search was performed on PubMed-Medline, Embase, and Cochrane Library by using key terms (osteonecrosis OR aseptic necrosis OR avascular necrosis AND core decompression).

Two independent reviewers (blinded by authors) screened the titles and abstracts. They also checked the reference lists of all potentially eligible studies and review papers to find out additional relevant publications. Among the searched publications, we selected studies, which met predefined inclusion and exclusion criteria. The inclusion criteria were (1) RCTs on clinical or radiologic results of CD in patients with ONFH and (2) studies reporting at least one of the following main outcomes: the rate of THA requirement, post-operative collapse rate, and clinical success. The exclusion criteria were (1) studies that included traumatic ONFH (2) non-RCT, reviews, protocols, basic science articles, and studies with less than ten subjects. The language was restricted to English. When updated studies, which involved the same cohort of patients in a previously published study, were identified, only the latest study was included in the analysis.

Types of interventions

Treatment modalities categorized into five; (1) traditional CD using 8–10-mm-wide cannula trephine and multiple drillings with a 3-mm Steinman pin, (2) CD combined with allogeneic or autologous BG, (3) CD combined with BMMC, (4) CD combined with BG and BMMC, and (5) non-operative treatments. Non-operative treatments included non-weight bearing with crutches, use of analgesics, physical therapy, and biophysical stimulation.

Data extraction and outcome measure

From every eligible study, we extracted the first author, year of publication, study design, number of patients, enrollment period, patient characteristics, type of treatment, and length of follow-up. The primary outcome of interest was the rate of conversion to THA. The secondary outcome of interest was the rates of radiographic progression and clinical success. We defined radiographic progression as a collapse of the necrotic femoral head more than 2 mm [13, 14].

Assessment quality and publication bias

Two authors (blinded by authors) independently performed quality assessment with use of the risk of bias assessment tool, which was described in the Cochrane Handbook for Systematic Reviews of Interventions [15]. We assessed publication bias with Begg’s funnel plot and Egger’s test.

Statistical analyses

Two statistical analyses were performed on the extracted data.

  1. 1)

    Pair-wise meta-analysis

First, we performed a pair-wise meta-analysis on comparative studies [16]. For each study, we calculated the relative risk with 95% confidence interval (CI) by using crude 2 × 2 tables, whenever possible. The Mantel–Haenszel method was used to calculate the odds ratio [17]. Heterogeneity between comparable studies was tested with chi-square (χ2) and I2 test. P > 0.1 and I2 < 50% were criteria of the statistical heterogeneity.

  1. 2)

    Network meta-analysis

After the pair-wise meta-analysis, we performed a NMA to determine comparative efficacy among the five treatments. The probability of superiority, which means the probability that a certain treatment has the highest success rates among the multiple treatments in simulation runs, was calculated. If a treatment had a 70% probability of superiority, it would mean that the treatment had the highest number of success rates out of all treatment modalities in 70% of the simulation runs.

All analyses were performed using the “mvmeta” command [18] of STATA (version 14.0; Stata Corporation, College Station, TX, USA) and self-programmed STATA routines described in Chaimani et al. [19] Corresponding 95% credible intervals (CrIs) were obtained using the 2.5th and 97.5th percentiles of the posterior distribution.

This study was exempted from institutional review board review because it did not involve human subjects.

Results

Description of included studies

We identified nine RCTs: two RTCs comparing CD and non-operative treatment [4, 5], four RTCs of CD versus CD with BMMC [20,21,22,23], two RTCs of CD with BG versus non-operative treatment [3, 24], and one RTC of CD with BG versus CD with BG and BMMC [14] (Figs. 1 and 2). A total of 453 hips were included in our NMA; 151 underwent CD, 70 CD with BG, 116 CD with BMMC, 25 CD with BG and BMMC, and 91 non-operative treatment. The characteristics of included studies are summarized in Table 1.

Fig. 1
figure 1

PRISMA flow diagram details the process of relevant clinical study selection

Full size image
Fig. 2
figure 2

Network plot depicting direct evidence used in network meta-analysis. Nodes representing the interventions being compared and edges representing the available direct comparisons. Nodes size and line thickness are weighted proportional to number of papers providing direct evidence

Full size image
Table 1 Characteristics of included individual studies
Full size table

Conversion to total hip arthroplasty

THA requirement data were provided in all nine studies. The overall rate of THA conversion was 24.2% (110/453 hips). There were no differences in the rate of THA conversion among the five treatments (Fig. 3a). The probability of superiority was 3.8% in CD, 1.5% in CD with BG, 64.5% in CD with BMMC, 24.5% in CD with BG and BMMC, and 5.8% in non-operative treatment.

Fig. 3
figure 3

The interval plot of the odds ratio for the efficacy by a the rate of conversion to total hip arthroplasty and b radiographic progression rate (black line indicates confidential interval; red line indicates credible interval). The 95% confidential interval included 1, and this means all treatment comparisons failed to reach a statistical significance

Full size image

Radiographic progression

Seven trials [3, 4, 14, 20, 22,23,24] provided data of radiographic progression. The overall rate of progression was 29.9% (116/387 hips). There were no differences in the rate of the femoral head collapse across the five treatments (Fig. 3b). The probability of superiority was 0.5% in CD, 1.1% in CD with BG, 21.4% in CD with BMMC, 64.1% in CD with BG and BMMC, and 13% non-operative treatment.

Clinical success

Four trials [3,4,5, 24] involving a total of 185 hips provided relevant data for clinical success while the remaining five studies described the change of pain scores [14, 20,21,22,23]. Thus, it was not feasible to perform NMA due to a lack of objective comparisons.

Methodological quality of included studies

Subjects were randomized by established allocation sequence of each study, and investigators were all blind to the allocation. Nevertheless, it was unclear whether the included trials met all the quality assessment criteria (Electronic Supplementary material 3), because physicians were not blind to their procedures.

Discussion

In our NMA, there were no significant differences in the rate of THA conversion and femoral head collapse among various CD modalities and non-operative treatment. During the last three decades, CD has been used in early-stage (pre-collapse) ONFH in an attempt to prevent future collapse of the necrotic femoral head, and the CD technique has evolved combining BG and BMMC. However, our results question the fundamental assumption that core decompression changes the natural course of ONFH.

Whether CD prevents future collapse of the necrotic femoral head remains an issue of controversy. While several studies reported excellent results of the procedure [4, 13, 23], other studies showed that it is not superior to non-operative treatments [6,7,8,9]. It is difficult to judge the efficacy of CD procedure due to the variety of modern techniques between studies, small number of subjects in each study and a paucity of RCTs. Moreover, the size of necrotic portion, which is the major determinant of future collapse of the necrotic femoral head, has not been stratified in most analyses. Most studies, which supported the efficacy of CD, indicated that the best results were obtained when this procedure was done in the earliest stage of the disease with small lesions [2, 25,26,27,28]. Considering that small lesions do not develop femoral head collapse even without any intervention, it should be argued that these studies might have had a selection bias including small early lesions.

The stage of ONFH also could be covariant factor in the analysis of the efficacy of CD. One review demonstrated that more recent studies have conferred better results than older studies, but there were fewer stage III hips in the more recent reports, suggesting that patient selection was an important reason for this improvement [26]. In our study, seven studies had included stage III lesions, so we expected that subgroup analysis based on the ONFH stage would reveal that patient selection was the main determinant of prognosis. However, subgroup NMA could not be performed, because only four studies identified their results according to the stage of disease (the other studies provided stage when enrolled but not at outcome evaluation) and there was too small sample size in each stage.

Although current study is the first NMA comparing conventional CD with recent modalities by summarizing all the available RCTs, our NMA still be interpreted cautiously. Because although various CD modalities have been categorized, the methods of surgical technique or adjuvant procedures within each category are not exactly the same.

Some limitations exist in our NMA study. First, the lesion size was not stratified. Second, the number of subjects in each study was small. Third, NMA includes indirect evidence, which necessitates another level of complexity and assumptions than pair-wise meta-analysis. Fourth, the preventive effect of CD in ONFH remains an issue of controversy. A large-scale RCT with stratification of necrotic extent is warranted to determine whether CD really prevents the disease progression in the patients with ONFH.