Background

Osteonecrosis of the femoral head (ONFH) is a common refractory disease in joint orthopedics. More than 10,000 new patients are affected with ONFH annually in the United States, accounting for approximately 10% of total hip arthroplasties (THAs) [1]. The cumulative number of patients with ONFH in China reached 8.12 million in 2013 [2]. According to statistics, the prevalence rate of ONFH is increasing yearly [3]. ONFH is a progressive disease typically caused by insufficient blood supply to the femoral head, which leads to increased pressure in it, eventually culminating in its collapse. The femoral head usually develops into secondary arthritis when it collapses [4]. Core decompression (CD) is a commonly used procedure for treating femoral head necrosis despite some controversy; it is a simple procedure that treats ONFH by drilling into the necrotic area of the femoral head [5,6,7,8]. The theoretical advantage of CD is in relieving the pain by reducing venous congestion and bone marrow pressure. Blood flow increases in the osteonecrosis area with the decrease in intraosseous pressure, thereby alleviating the pathology and promoting bone regeneration in the osteonecrosis area [9, 10]. CD combined with regeneration therapy appears to accelerate the healing of osteonecrosis and reduce the risk of femoral head collapse [11]. Recently, studies have shown that bone marrow aspirate concentrate (BMAC), expanded mesenchymal stem cells (MSCs), autologous bone graft (ABG), and other regenerative therapies show gratifying outcomes in the treatment of bone diseases [12,13,14,15,16,17]. In addition to BMAC, MSCs, and ABG, common regeneration therapies also include platelet-rich plasma (PRP), autologous bone graft combined with bone marrow aspirate concentrate (ABG + BMAC), and free vascular autologous bone graft (FVBG). Despite the promising results of these diferent methods, the best regeneration therapy for ONFH has not yet been determined.

Bayesian network meta-analysis (NMA), also known as multiple treatment comparison meta-analysis, can simultaneously analyze direct and indirect evidence from different studies, expand the scope of traditional conventional pairwise analysis, and subsequently estimate the relative effectiveness of all interventions and rank them [18]. To date, no comparison of the different regenerative therapies has been performed for ONFH using NMA. Herein, we used a Bayesian NMA to evaluate the efficacy of different regenerative therapies based on ONFH progression and conversion to THA.

Methods

This systematic review and NMA adhered to the guidelines outlined in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [19]. Additionally, our review was registered on PROSPERO (http://www.crd.york.ac.uk/PROSPERO) under the registration number CRD42023412784.

Search strategy

All articles published between 2003 and 2023 in PubMed, EMBASE, Cochrane Library, and Web of Science databases were searched. We used the following keywords: “femur head” AND (“bone necrosis” OR “avascular necrosis” OR “osteonecrosis”) AND (“regenerative therapies” OR “stem cells” OR “bone marrow” OR “bone graft” OR “platelet rich plasma”). An additional file presents the details of the search process (see Additional file 1).

Study selection

The inclusion and exclusion process followed the PICOS (Participants, Intervention, Comparison, Outcome, and Study) principle. Additionally, the mean age of patients with ONFH was 18 years. Studies include at least two of the following treatments: ABG, ABG + BMAC, BMAC, CD, MSCs, FVBG, and PRP. The included studies reported at least one of the two outcomes as follows: the rate of THA requirement and that of ONFH stage progression after the intervention. Furthermore, the included studies were randomized controlled trials (RCTs) or retrospective cohort studies conducted in English, published from 2003 to 2023.

The exclusion criteria were as follows: Non-English text; literature with a low-quality treatment evaluation; and reviews, protocols, case reports, conference papers, and animal experiments.

All relevant studies were screened independently by two reviewers, and any disagreement between the two reviewers regarding a study’s eligibility was resolved through discussion with a third reviewer.

Data extraction

Two independent reviewers extracted the following information from each included study: the first author’s surname, year of publication, study types, follow-up time, average age, hip sample size, ONFH staging of patient, conversion to THA, and ONFH progression. Any differences were resolved through discussion with a third reviewer.

Quality assessment

Two independent reviewers assessed the literature quality. RCT and retrospective cohort studies were assessed for quality using the Cochrane Risk of Bias tool [20] and the Newcastle–Ottawa Scale (NOS), respectively. The following factors were assessed for each study: randomization sequence generation (selection bias), allocation concealment, subject blinding, outcome assessment, attrition bias (incomplete outcome data), reporting bias (selective reporting), and other biases. Studies with scores of 8 and 9, and 6 and 7 were considered high and medium quality studies, respectively [21]. Any differences were resolved through discussion with a third reviewer.

Statistical analysis

We analyzed the following two metrics: ONFH conversion rates to THA and its progression rates. The results are expressed as the odds ratio (OR) and 95% confidence interval (CI). A pairwise meta-analysis was performed using R software (version 5.35; Lucent Technologies, Paris, France).

Heterogeneity between comparable studies was examined using the chi-square (χ2) and I2 tests. Values < 25%, 25–75%, and > 75% for the I2 statistic represented mild, moderate, and severe heterogeneity, respectively [22]. Furthermore, node-splitting analysis was used to assess the inconsistency of a particular comparison based on direct and indirect evidence; statistical significance was considered at P < 0.05 [23]. Furthermore, funnel plots were used to test for publication bias.

We also calculated the surface under the cumulative ranking (SUCRA) value, a simple numerical summary to supplement the graphical display of cumulative ranking, which is used to estimate the SUCRA line for each treatment. The SUCRA values of 1 and 0 signify a treatment that is certain to be optimal and the worst, respectively [23].

Results

Study selection and characteristics

Figure 1 shows the study selection process. We retrieved 2591 articles, ultimately including 17 studies. A total of 1019 hips were included in our NMA groups, comprising 245, 80, 25, 177, 50, 151, and 291 hips in the BMAC, MSCs, PRP, ABG, ABG + BMAC, FVBG, and CD groups, respectively. Table 1 presents the basic characteristics of the included studies. The network structure of the analyzed comparisons for the primary outcomes is shown in Fig. 2.

Fig. 1
figure 1

Flowchart of study selection and design

Full size image
Table 1 Characteristics of included individual studies
Full size table
Fig. 2
figure 2

Network plots of comparison-based network meta-analyses. Each circular node represents a type of intervention. The circle size is proportional to the total number of patients. The width of the lines is proportional to the number of studies performing head-to-head comparisons in the same study. a ONFH progression and b conversion to total hip arthroplasty (THA). ABG: autologous bone grafting; ABG + BMAC: autologous bone grafting and bone marrow aspirate concentrate; BMAC: bone marrow aspirate concentration; CD: core decompression; FVBG: free vascular vascularized bone grafting; MSCs: mesenchymal stem cells; PRP: platelet-rich plasma

Full size image

For RCTs using the Cochrane Risk of Bias tool, the overall quality assessment showed a low or moderate risk of bias, with a higher risk observed in the blinded component, mainly because the procedure required informed consent and it was difficult for the operator and patient to be blinded; however, this does not imply that the study was meaningless (Fig. 3). Retrospective cohort studies were assessed using the NOS, revealing two medium-quality (score: 6 or 7) and two high-quality (score: 8 or 9) studies (Fig. 4). We assessed a funnel diagram of the included studies (Fig. 5), and the roughly symmetrical diagram suggests no publication bias.

Fig. 3
figure 3

Assessing the quality of randomized control trials (RCTs) using the Cochrane Risk of Bias tool, a: Risk of bias graph. b: Summary of study risk bias analysis

Full size image
Fig. 4
figure 4

Assessing the quality of retrospective cohort studies using the Newcastle–Ottawa Scale

Full size image
Fig. 5
figure 5

Funnel diagram of the included studies. a Osteonecrosis of the femoral head (ONFH) progression and b conversion to total hip arthroplasty (THA)

Full size image

Heterogeneity was found among the comparisons of treatments on the rates of ONFH progression and conversion to THA (Tables 2, 3). We found that the Bayesian NMA results were reliable; the inconsistency between the direct and indirect effects of comparisons of different treatments on the two outcomes showed no significant differences (Tables 4, 5). The nodal split method for ONFH progression and conversion to THA showed no significant heterogeneity (Tables 6, 7).

Table 2 Heterogeneity results of ONFH progression according to pairwise meta-analysis
Full size table
Table 3 Heterogeneity results of conversion to THA according to pairwise meta-analysis
Full size table
Table 4 Odds ratios of osteonecrosis of the femoral head progress between treatment groups
Full size table
Table 5 Odds ratios of conversion to total hip arthroplasty between treatment groups
Full size table
Table 6 The results of node-splitting method for ONFH progression
Full size table
Table 7 The results of node-splitting method for conversion to THA
Full size table

Femoral head necrosis progress

All 17 articles reported ONFH progression. The NMA results showed that MSCs (OR: 0.098, 95% CI: 0.0087–0.87, SUCRA = 0.705) were the first effective technique for preventing ONFH progression, followed by BMAC (OR: 0.27, 95% CI: 0.073–0.73, SUCRA = 0.431). However, no significant difference was found among VBG, PRP, ABG + BMAC, ABG, or CD in preventing ONFH progression (P > 0.05) (Fig. 6).

Fig. 6
figure 6

Forest plots of osteonecrosis of the femoral head (ONFH) progression. ABG: autologous bone grafting; ABG + BMAC: autologous bone grafting and bone marrow aspirate concentrate; BMAC: bone marrow aspirate concentration; CD: core decompression; CI, confidence interval; FVBG: free vascular vascularized bone grafting; MSCs: mesenchymal stem cells; PRP: platelet-rich plasma

Full size image

Conversion to total hip arthroplasty

A total of 16 articles reported ONFH conversion to THA. MSCs (OR: 0.062, 95% CI: 0.0038–0.40, SUCRA = 0.902) and BMAC (OR: 0.32, 95% CI: 0.1–0.074, SUCRA = 0.511) were the first and second effective techniques, respectively, for preventing ONFH conversion to THA. No significant difference was found among VBG, PRP, ABG + BMAC, ABG, or CD in preventing ONFH conversion to THA (P > 0.05) (Fig. 7).

Fig. 7
figure 7

Forest plots of conversion to total hip arthroplasty (THA). ABG: autologous bone grafting; ABG + BMAC: autologous bone grafting and bone marrow aspirate concentrate; BMAC: bone marrow aspirate concentration; CD: core decompression; CI, confidence interval; FVBG: free vascular vascularized bone grafting; MSCs: mesenchymal stem cells; PRP: platelet-rich plasma

Full size image

Discussion

This NMA investigated regenerative therapy for nontraumatic femoral head necrosis and included data from 17 clinical trials, including 1019 hips assigned to 6 different treatment options. The quality of evidence was generally low or moderate risk of bias. Our NMA found that MSCs and BMAC can prevent ONFH progression and conversion to THA. When ranked according to the SUCRA value, MSCs were the first, followed by BMAC; therefore, we concluded that MSCs and BMAC transplantation may be necessary after CD. To our knowledge, this is the first NMA to compare these six regenerative therapies, and these findings provide evidence for regenerative therapy for ONFH.

Previous traditional meta-analyses, such as the study by Andriolo et al. [11], did not compare different regenerative therapies separately with CD but combined the data of different regenerative therapies, which may have led to biased results. Zhang et al. [49] found that the combination of bone marrow stem cells had better prognosis outcomes than CD alone, such as ONFH progression or Harris Hip Score. However, during their study search, they classified bone and PRP graftings as stem cells, which broadened the study’s scope but inevitably increased its bias. Our study categorized regenerative therapy into six categories, which improved the search accuracy and obtained reliable study results. Migliorini et al. [50] found that bone marrow-derived cells had a lower probability of THA than CD, whereas conventional meta-analyses only compared bone marrow-derived cells with CD alone and could not compare multiple regenerative therapies. Bayesian NMA was used to review the regenerative therapy for ONFH in the CD (control), ABG, ABG + BMAC, BMAC, FVBG, MSCs, and PRP groups. Based on our NMA SUCRA analysis, MSCs ranked as the first intervention among the six regenerative therapies for preventing ONFH progression and conversion to THA, presumably because they provide better repair capacity [17, 39, 47]. However, ABG, ABG + BMAC, FVBG, PRP, and CD showed no significant differences in preventing ONFH progression and conversion to THA.

We derive the rationale behind this conclusion from the premise that regeneration therapy operates on the ability of cells and molecules to induce and promote tissue repair of ONFH. For example, BMAC contains many growth factors and non-mesenchymal cells, including endothelial, hematopoietic, and inflammatory cells, in which growth factors can induce stem cells to migrate to the injured site [40,41,42]. Many hematopoietic stem cells can provide vascular support and drive MSCs toward osteogenic differentiation [43]. With the progress in BMAC research, researchers have found that the non-progenitor cell component of BMAC may negatively affect its regeneration characteristics, limiting BMAC’s repair ability [44]. Generally, MSCs account for only 0.001–0.01% of the number of nucleated cells in BMAC [45, 46]. Studies have shown that high concentrations of MSCs can promote cartilage healing more than low concentrations of MSCs without causing adverse reactions [47]. Additionally, MSCs can differentiate into several cell types (including fibroblasts, chondroblasts, and other forms of tissue regeneration cells), thereby promoting tissue repair [48].

Although our study is the first Bayesian NMA to compare traditional CD with other regenerative therapies, it has limitations. First, it included only 17 related articles; therefore, the scale of direct comparison was limited. For example, only 80 and 245 hips were in the MSCs and BMAC groups, respectively; therefore, a limited sample size may increase statistical dispersion. Second, the study’s sample size was not large enough, potentially reducing the credibility of the results. Moreover, including patients with different Association Research Circulation Osseous (ARCO) stages leads to heterogeneity, and the prognoses of patients with different stages may differ. Considering these limitations, we recommend caution in our conclusions. Therefore, future research studies should include larger sample sizes covering the various treatments and ARCO/Ficat stages.

Conclusion

Our NMA found that MSCs and BMAC were effective in preventing ONFH progression and conversion to THA among the six regenerative therapies. MSCs ranked first, followed by BMAC according to the SUCRA value. Based on our NMA results, MSCs and BMAC after CD may be necessary to prevent ONFH progression and conversion to THA. Furthermore, these findings provide evidence for regenerative therapy for ONFH.