Introduction

FNF is a serious traumatic disease, accounting for about 50% of hip fractures. It mainly occurs after falls in the elderly, and is more common in middle-aged people due to violent injuries [1]. In the United States, the incidence of FNF is 63.3 per million in women and 277 per million in men [2]. Although detailed epidemiological data yet have not been available in China, it is estimated that a significant number of this group suffer from FNF. With the aging population increasing and the social modernization forging ahead, the incidence of FNF is on the rise in this country.

Though various treatments for FNF have sprung up, IF with autologous hip joint is still the mainstream [3,4,5]. ONFH, or avascular necrosis of femoral head, is the ultimate consequence of impaired blood supply. ONFH and nonunion are two major complications after IFFNF, especially in the young group [6, 7]. As the technology of IF advances, the healing rate of FNF has remarkably improved [8,9,10]. However, recent years have witnessed a nonsignificant drop of the occurrence of ONFH after IFFNF [6, 11].

The risk factors of ONFH after IFFNF have been reported by many lately, but the results of studies are divergent and even controversial. In this study, we retrieved the published literature on ONFH after IFFNF, extracted high-relevant data for a systematic review and meta-analysis, and evaluated the significance of the filtered risk factors, in a bid to provide a reference for the prevention and treatment of this complication.

Methods

Search strategies

This study was executed in line with the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [12] and reported based on the guidelines developed by the Meta-Analysis of Observational Studies in Epidemiology group [13]. Because all the analyses were performed on the basis of previous published studies, no ethical approval or patient consent was required. In the initial screening, 2 investigators (J-LX and B-LX) conducted the main search in the electronic databases of PubMed, Embase and Cochrane Library to retrieve eligible articles on ONFH-related risk factors after IFFNF from the inception of the databases to June 2019, without restrictions to languages, publication types or regions. The combined terms of Medical Subject Headings (MeSH) and non-MeSH were searched as follows: femoral neck fractures, femoral neck fracture, femur neck fractures, femur neck fracture, femur head necrosis, femur head necroses, head necrosis femur, necrosis femur head, aseptic necrosis of femur head, necrosis aseptic of femur head, necrosis avascular of femur head, ischemic necrosis of femoral head, avascular necrosis of femoral head, primary avascular necrosis of femur head,fracture fixation, internal, and internal fixation. A third investigator irrelevant to the initial procedure was consulted in case of any discrepancy.Taking the pubmed database as an example, the literature search strategy is shown in Table 1.

Table 1 The Pubmed database literature search strategy
Full size table

Study selection criteria

Two independent investigators (J-LX and Z-RL) analyzed the initially selected articles to verify their relevance with the topic of risk factors and ONFH after IFFNF. Studies had to fulfill the following criteria for inclusion: outcome was femoral head necrosis;internal fixation for femoral neck fracture;study design included case-control, retrospective,and prospective cohorts, and cross-sectional studies;participants were selected without limitations to regions, ages or social status.Trials were excluded according to following identifications: duplicate or overlapping data, animal experiments, conference abstracts, letters and review articles. In case of any disagreement the results were discussed and unified by senior authors.

Data extraction

Data from the included studies were extracted and independently categorized by 2 of the authors (J-LX and Q-ZZ) in a predefined data extraction form. All disagreements were resolved by discussion. Design information, baseline population characteristics (mean age, sample size and country), surgical approach, risk factors from all included studies were stratified into a standardized evidence table. All the data were rechecked to ensure accuracy. Study selections were shown in a PRISMA flow diagram.

Methodological quality assessment

The methodological quality of the included studies was evaluated by 2 independent reviewers (J-LX and T-YL) based on the items of modified Newcastle-Ottawa Scale (NOS) [14], comprising patient selection, study group comparability and outcome assessment. The observational studies scored 0 to 9. Divaricate opinions were discussed among the authors.

Statistical analysis

The meta-analysis and statistical analysis were performed using Cochrane Collaboration Review Manager software (RevMan version 5.3, Nordic Cochrane center, Copenhagen, Denmark). The odds ratio (OR) was utilized as a statistical analysis. The I-square (I2) test was adopted to evaluate the influence of heterogeneity on the output of meta-analysis. I2 values of 0, 25, 50 and 75% represented no, low, medium and high heterogeneity, respectively. According to the Cochrane review guidelines, severe heterogeneity of I2 ≥ 50% required the utilization of random-effect models. Otherwise, the fixed effect model was approved. P value less than 0.05 was accepted as statistical significance. Sensitivity analysis [15] was conducted by study removal approach to evaluate the quality and consistency of the results. Funnel plots were visually checked, and Egger and Begg linear regression tests of publication bias were carried out by Stata 13.0 software.

Results

Study selection process

As a result, 765 references were initially retrieved, 740 were left after eliminating duplicate literature; and then 716 without high-relevant to our topic were discarded by reading titles and abstracts, and 24 studies remained. Finally, 7 full-text articles were abandoned because of the following reasons: 2 studies did not provide with full texts; 5 studies lacked the independent groups of ONFH. Therefore, 17 observational studies with 2065 patients were included in the meta-analysis. The flow chart describing the selection process of the study was shown in Fig. 1.

Fig. 1
figure 1

The flow diagram of literature selection

Full size image

Study characteristics and methodological quality

The 17 included references were case-control studies, with the publication years differing from 2005 to 2019. Nine were conducted in China (including 1 in Taiwan), 2 in South Korea, and 1 in Malaysia, Chile, France, Saudi Arabia, UK and Brazil, respectively. In the selected clinical trials, the sample sizes varied between 29 and 277 participants.12 studies reported cannulated screw internal fixation, 3 studies reported cannulated screw fixation or dynamic hip screw fixation, 1 study reported dynamic hip screw fixation, and 1 study reported cannulated screw fixation or cannulated screw combined with dynamic hip screw fixation or locking compression plate fixation.2 studies reported preoperative traction,1 study reported skeletal traction, and another study did not indicate traction type;3 studies reported the status of implants,including removal and retention of implants in the body.2 studies reported the mechanism of injury,including high-energy injuries and low energy injuries;3 studies reported fracture reduction mode,including open reduction and closed reduction.The basic characteristics of the 12 of them were shown in Table 2. In addition, all the studies were evaluated as high methodological quality in accordance with the NOS.

Table 2 Characteristics of the Included Studies
Full size table

Risk factors of ONFH after IFFNF

Gender

Of the 16 included studies, 4 reported [16, 19, 26, 33] the correlation between gender and ONFH after IFFNF. Because of nonsignificant heterogeneity among the studies, the fixed effect model was utilized (I2 = 25%, P = 0.26). However, nonsignificant differences in gender were unveiled between ONFH patients and normal ones (OR: 1.27, 95%CI: 0.84–1.94, P = 0.26). (Fig. 2).

Fig. 2
figure 2

Meta-analysis results of the correlation between gender and ONFH after IFFNF.

Full size image

Age

Data on age were available for the meta-analysis from 4 studies [16, 17, 19, 28], and nonsignificant heterogeneity was presented among the studies (I2 = 17%, P = 0.31). Therefore, the fixed effect model was applied. However, differences in age between ONFH patients and normal ones were still nonsignificant (OR: 1.66, 95%CI: 0.89–3.11, P = 0.11). (Fig. 3).

Fig. 3
figure 3

Meta-analysis results of the correlation between age and ONFH after IFFNF.

Full size image

Garden classification

Data on Garden classification were available for the systematic analysis from 14 studies [16,17,18,19,20, 22,23,24, 26,27,28,29,30, 32]. The fixed effect model was chosen due to nonsignificant heterogeneity in intra-study comparisons (I2 = 1%, P = 0.44). The outcome manifested a statistically significant difference in the item between the two patient groups (OR: 0.40, 95%CI: 0.29–0.55, P < 0.00001). (Fig. 4).

Fig. 4
figure 4

Meta-analysis results of the correlation between Garden classification and ONFH after IFFNF.

Full size image

Injury-operation interval

Five references [19, 20, 26,27,28] concerning the data on injury-operation interval were available for the analysis. The fixed effect model was used as nonsignificant heterogeneity was found in intra-study comparisons (I2 = 36%, P = 0.18). No statistically significant difference in this index was shown between the two patient groups (OR: 1.29, 95%CI: 0.82–2.04, P = 0.27). (Fig. 5).

Fig. 5
figure 5

Meta-analysis results of the correlation between injury-operation interval and ONFH after IFFNF.

Full size image

Fracture reduction mode

Three studies [16, 21, 26] containing statistics on fracture reduction mode were available for the analysis using the fixed effect model, without significant heterogeneity among the studies (I2 = 33%, P = 0.23). The results exhibited no statistically significant differences in fracture reduction mode between the two patient groups (OR:1.98, 95%CI: 0.92–4.26, P = 0.08). (Fig. 6).

Fig. 6
figure 6

Meta-analysis results of the correlation between fracture reduction mode and ONFH after IFFNF.

Full size image

Preoperative traction

Information in 2 [16, 19] of the references over preoperative traction was included in meta-analysis by the random effect model due to a significant statistical heterogeneity (I2 = 81%, P = 0.02). The results shown that there was nonsignificant differences in preoperative traction between ONFH patients and normal ones (OR:1.69, 95%CI: 0.29–9.98, P = 0.56). (Fig. 7).

Fig. 7
figure 7

Meta-analysis results of the correlation between preoperative traction and ONFH after IFFNF.

Full size image

Status of implants

Three studies [16, 19, 23] reporting statistics on the status of implants were involved in meta-analysis. There was no significant statistical heterogeneity among the studies (I2 = 25%, P = 0.26) and the fixed effect model was utilized. It was found that the difference in implants status was significant between the two patient groups (OR: 0.04, 95%CI: 0.02–0.07, P < 0.00001). (Fig. 8).

Fig. 8
figure 8

Meta-analysis results of the correlation between status of implants and ONFH after IFFNF.

Full size image

Mechanism of injury

Two references [19, 21] explored the underlying mechanism of the injury. The relevant data were analyzed by the random effect model, for a significant statistical heterogeneity was found among the studies (I2 = 80%, P = 0.03). There was a significant difference in the mechanism between ONFH patients and normal ones (OR: 0.53, 95%CI: 0.06–4.83, P = 0.58). (Fig. 9).

Fig. 9
figure 9

Meta-analysis results of the correlation between mechanism of injury and ONFH after IFFNF.

Full size image

Sensitivity analyses

The sensitivity analysis was performed on the selected studies to assess whether individual studies would affect the overall results. The outcomes suggested that data of one study [25] strongly affected the overall results (Fig. 10). Rereading the article verbatim, we found that the study actually accepted 150 cases with ONFH of 325 in total; and the incidence was 46.15%, much larger than the others, averaged 22.95%. Therefore, we discarded this reference from the meta-analysis.

Fig. 10
figure 10

Influence analysis of included studies

Full size image

Evaluation of publication bias

Visual inspection of funnel plots was adopted in the estimation. Specifically, Egger and Begg tests uncovered publication bias in this study (P = 0.01). This may attribute to incomplete retrieval or unpublished negative results. (Figs. 11, 12 and 13).

Fig. 11
figure 11

Funnel plot of correlation between Garden classification and ONFH after IFFNF.

Full size image
Fig. 12
figure 12

Egger’s funnel plot

Full size image
Fig. 13
figure 13

Begg’s funnel plot

Full size image

Discussion

In this study, we conducted a meta-analysis of 16 selected studies to corroborate the risk factors of ONFH after IFFNF. To ensure a reliable conclusion, previous published studies on ONFH after IFFNF were retrieved, reviewed and summarized to achieve those with high compliance and high quality, so as to answer various clinical questions of this malady. Overall, our results suggest that Garden analysis and the retention of internal fixators are critical risk factors of ONFH after IFFNF. Gender, age, injury-operation interval, fracture reduction method, preoperative traction and mechanism of ONFH cannot be considered as correlative factors of ONFH after IFFNF. In addition, the publication bias was found by the visual distribution of funnel plots and Egger and Begg tests, indicating incomplete retrieval or unpublished negative results. The sensitivity analysis indicated one research, by Zeng XS et al [25], had significant influence on the overall results, so the study was excluded.

Orthopaedic surgeons undoubtedly encounter great challenges on the treatments for FNF, for contentions between distinct therapies have persisted for years. Generally, the commonly acceptable treatment is IF by open or closed reduction for young patients or ones without degenerative changes in the hip. Total hip arthroplasty is usually the preferable option for patients with extant hip degenerative changes. The primary purpose of IF for the fractures under this condition is to achieve anatomic reduction, which is salutary to restore or maintain the unstable blood supply to the femoral head and prevent complications (e.g., ischemic necrosis and non-healing) in the days to come.

Garden classification is considered as an important factor in the formation of ONFH after IFFNF. Strong relationship is reported between the risks of avascular necrosis in displaced intracapsular fractures and undisplaced FNF [33,34,35]. Sun et al [36] suggested that screw removal might be a primary inducing factor in the pathogenesis of ONFH, because all compressive, tensile and shear stresses were concentrated on the fracture site after screw removal, changing the biological stress thereby. Yao S et al [37] suggested that the intraosseous pressure of the femoral head could increase in that the presence of the screw exacerbated the femoral head ischemia.

As the passages have exposed, two significant advantages of our study are clear. Firstly, as the risk factors of ONFH after IFFNF were controversial, this meta-analysis assessed the potential correlation between ONFH and ten more indexes (gender, age, Garden classification, etc.) after IFFNF through a thorough systematic study with rigorous analytical methods. Secondly, the rationality and reliability of our meta-analysis have been prudently and significantly improved in that the overall comprehensive estimation is based on a large sample size. In addition, sufficient sensitivity analysis has been carried out to ensure the reliability of this study.

The current meta-analysis has the following limitations and must be considered before our results are accepted. First, the selected studies in the meta-analysis were published between 2005 and 2019. The influencing factors reported in the studies were not completely consistent, and most of them did not adjust for confounding factors. Second, the values of meta-analysis in this study were derived from univariate analysis, with residual confounding and unmeasured factors. Our more accurate results will come from adjustments for other confounding factors. Third, the research included in this analysis is insufficient, and potential publication bias still exists. Fourth, this study only includes references in English. Therefore, we may have lost data from those in other languages.

Conclusion

In summary, our meta-analysis suggests that Garden classification and the retention of IF are important factors affecting ONFH after IFFNF. However, because of the publication bias of this study and the results from the univariate analysis, we cannot determine whether the two items are independent risk factors of ONFH after IFFNF. In this study, gender, age, injury-operation interval, fracture reduction mode, preoperative traction and the mechanism of ONFH have been ruled out from the consideration to be correlative factors with this complication after IFFNF.