Autologous bone graft in the treatment of post-traumatic bone defects: a systematic review and meta-analysis

Background This meta-analysis aimed to determine the bone union rate of bone defects treated with the different autologous bone graft techniques. Methods The PubMed and the Cochrane Library databases were searched using the terms: ‘fracture’ AND (‘bone loss’ OR ‘defect’ OR ‘defects’) AND ‘bone graft’, restricted to English language, to human species, and to a publication period from January 1999 to November 2014. Data were extracted by one of the reviewers and then checked by the second. A quality of evidence score and a methodology score were used. Heterogeneity was assessed. A random effects model approach was used to combine estimates. Results Out of 376 selected studies only 34 met the inclusion criteria. The summary pooled union rate was 91 % (95 % CI: 87–95 %) while union rate after additional procedures raised to 98 % (95 % CI 96–99 %). No association between union rate and bone defect size was found. (Univariable regression model: vascularized: P = 0.677; non-vascularized: 0.202. Multivariable regression model: vascularized: P = 0.381; non-vascularized: P = 0.226). Vascularized graft was associated with a lower risk of infection after surgery when compared to non-vascularized graft (95 % CI 0.03 to 0.23, p < 0.001). Conclusion The results of this meta-analysis demonstrate the effectiveness of autologous graft for bone defects. Furthermore, from the available clinical evidence bone defect size does not seem to have an impact on bone union when treated with autologous bone graft techniques. Electronic supplementary material The online version of this article (doi:10.1186/s12891-016-1312-4) contains supplementary material, which is available to authorized users.


Background
Conventional autologous bone graft has become the most widely used treatment for bone defects over time. Several factors contributed to its widespread application: it is easy to obtain, it combines osteogenic, osteoinductive and osteoconductive properties, it does not raise immune response or transmit infectious diseases [1,2]. Furthermore, autologous bone graft can be harvested in a variety of forms and sizes from different donor sites [1,2].
Selection of the autologous graft type in the treatment of bone defects has been mostly based on defect size: several authors do not recommend the use of the non-vascularized graft in defects larger than 5 cm [1][2][3]. The more technical demanding vascularized bone graft method is considered the best choice for larger size defects [2,3]. However, in recent years, advances in graft harvesting technique [4] and in wound environment recovery using the polymethylmethacrylate (PMMA) induced membrane technique [5], renewed the interest in the use of the non-vascularized autologous bone graft. Infection also plays a role in graft selection and a 2-stage approach with delayed grafting is sometimes necessary [6].
The primary objective of this meta-analysis is to determine the bone union rate of post-traumatic bone defects treated with the different autologous bone graft techniques. The secondary objective is to determine the rate of infection after this treatment.

Data collection and extraction
Prior to doing the electronic search, a written protocol was established according to guidelines for systematic reviews (AMSTAR, MOOSE and PRISMA) [7][8][9]. An electronic search was conducted in Medline restricted to English language, to human species, and to a publication period from January 1999 to November 2014. The search terms and Boolean operators used were: 'fracture' AND ('bone loss' OR 'defect' OR 'defects') AND 'bone graft'. Additionally an electronic search was done in the Cochrane Library with the terms: fracture AND bone loss AND defect OR defects AND bone graft.
Two reviewers (MA, AA) independently scrutinized the list of titles of all the retrieved citations and, if necessary, the abstracts to determine usefulness of the article. The final selection was based on the full text version of the potentially relevant articles that were assessed independently by the reviewers. All references cited in these elected studies were manually searched along with the "related articles" researches in PubMed engine for additional relevant studies. Papers published by the same research group and studying the same factors were checked for duplicate data. Where duplication occured the less detailed paper was discarded.
We included only original reports that presented the results of at least ten cases of bone defects secondary to open fractures, post-traumatic nonunion or infected bone resection. The exclusion criteria were: bone defects after tumor resection; bone defects after reduction and fixation of closed metaphyseal fracture of long bones; studies with more than 25 % of the defect not located in long bones (forearm, humerus, femur or tibia); studies with more than 25 % of the defects treated with osteoconductive biomaterials in addition to the bone graft; cases with the use of osteoinductive factors in the graft; studies mostly about bone defects in children and studies that did not report the information about healing after treatment. When the information of each patient in a study was presented in the text and/or tables, cases that met the exclusion criteria were removed and the remaining patients were enrolled in the analysis.
Included studies were classified according the Oxford Centre for Evidence-Based Medicine system and a modified version of the Coleman methodology score [10] (Additional file 1). Data was extracted by one of the reviewers and then checked by the second. Disagreements were solved via discussion and consensus between the two reviewers. The following definitions were used for data extraction: primary union described as bone union achieved after bone grafting, secondary union as bone union achieved with a further surgery after the bone graft. Of note, a graft fracture was considered a union related complication only when the original study classified it in this manner, and cases with union before lost of follow up were considered as treated. Treatment failures were viewed as the loss of the graft in the postoperative period that required debridement and a new graft, the absence of bone union during follow-up or a new bone defect treatment (bone transport, amputation, etc.). Preoperative infection refers to the presence of infection (active or quiescent) or absence of it when bone defect treatment was implemented. Postoperative infection was infection reported as a complication after bone graft procedure. We considered that PMMA was used as an adjuvant in bone defect treatment (induced membrane technique) only when authors reported its use for this purpose.

Assessment of publication bias
Susceptibility of the systematic review to publication bias was formally assessed with the Egger test [11].

Quantitative data synthesis
To stabilize variance, the bone union proportions were subject to a Freeman-Tukey arcsine square root transformation and back-transformed according to Miller after quantitative data synthesis [12,13]. With the normalized data, heterogeneity was assessed using both Cochran's Q test and the inconsistency measure I 2 suggested by Higgins [14]. A cut-off of P < 0.10 was used to indicate heterogeneity. Values of I 2 equal to 25 %, 50 % and 75 % denoted a low, moderate and high degree of statistical heterogeneity. As data from a series of studies that had been performed independently are thought to be not functionally equivalent, a random effects model approach was used to combine estimates. Confidence intervals within studies were achieved using the exact binomial method. To perform a sub-group analysis, the studies were divided according to graft vascularization in two major categories: non-vascularized bone graft or vascularized bone graft. Analyses were performed using STATA (version 13.0) and Comprehensive Meta-analysis (version 2.0).

Selection of studies
The Medline search resulted in 338 citations and after the abstract review 21 were considered as potentially eligible and all of them had the full version reviewed. References of these articles were manually screened and also the related citations tool resulting in further 38 potentially eligible articles, totaling 59 papers to review. The Cochrane Library search did not result in additional studies. After the full version review, 34 studies met the inclusion/exclusion criteria ( Fig. 1) (Additional file 2). In seven of the 34 studies some cases were excluded from the analysis (Additional file 3). A total of 749 patients with 750 bone defects were included in this meta-analysis.
Concerning study characteristics, one was a randomized controlled trial, one was a prospective case series, three were retrospective comparative case series and 29 were retrospective case series (Table 1). Only Pelissier et al. [15] was a comparison between vascularized and non vascularized bone graft. The studies achieved 37.1 points (21 to 72) out of 100 in the quality assessment tool. The inter rater agreement in regards to the quality assessment between the reviewers was considerably high (ICC = 0.78; 95 % CI 0.75 to 0.94).

Publication bias
The shape of the funnel plot revealed evidence of asymmetry for both primary and secondary union (Fig. 2). The Egger's test showed evidence of publication bias (p < 0.001 for primary union and p < 0.001 for secondary union).
For comparison of vascularized versus non-vascularized graft the study from Toh et al. [18] and the study from Muramatsu et al. [20] were omitted as they used a mixed technique. The study of Pelissier et al. [15] included patients treated either with vascularized or non-vascularized graft, thus this publication contributed to both groups in the sub-group analysis ( Table 2).
Additionally, investigation of heterogeneity was performed by means of meta-regression including age, percentage of females and months of delay from injury to treatment. No variables showed a significant association with the risk of post-treatment infection.

Bone union
The primary bone union rate expected for the bone graft techniques is 91 %. In some circumstances, additional procedures such as the change of a broken implant, compression in the nonunion site or cancellous graft in nonunion areas at bone ends, may be necessary and they raised the union rate to 98 % in published studies (Additional file 4) [15, 17-23, 25-31, 33-35, 39, 41, 42, 46-48].

Defect size as a guide to select graft
Only few studies presented a description of the method used to define and measure the bone defect [27,29,31,46]. Small defects that might have been susceptible to spontaneous regeneration were present in some studies. They were treated not only with non-vascularized graft but also with vascularized bone graft.
Studies about vascularized bone grafts have been performed on larger bone defects but association of union rate and bone defect size wasn't found between the vascularized and non-vascularized grafts. Limitations of this conclusion include also a potential selection bias: some recent studies about non-vascularized graft were excluded because of the addition of growth factor or biomaterial to the graft. Despite the limitation of this study, our data suggests that selection of graft technique shall not be guided only by defect size. Patient expectations, surgeon experience, soft tissue condition and a trained staff to perform microsurgery are elements that must be carefully judged before making a decision on the graft to be used. The pooled estimate of mean effect size showed a decrease of infection after treatment compared with the pre-operative situation. However, these findings should be interpreted with caution due to the presence of a moderate degree of statistical heterogeneity. According  Again, this conclusion is limited. Infection definition varies between the included studies and several different surgical techniques were used. Although we cannot give evidence to support this recommendation, most of the studies suggest a two step reconstruction as the standard approach to manage infected bone defects: an extensive debridement, followed by antibiotic treatment before graft surgery [16,21,26,29,31,32,36,40,43,44]. Furthermore, some of the studies use PMMA as a local antibiotic delivery and/or due to its ability to induce a biological membrane at the defect site [21,29,[32][33][34]36].

Overall completeness and applicability of evidence
The included studies provide the most complete information available concerning union rates after autologous graft for bone defects; however, different factors may have added to the heterogeneity of the pooled results, such as different treatment techniques, different sample sizes reflecting different levels of experience, incomplete information about complications. Additionally, information regarding surgical steps was limited in several studies. Finally, data concerning potential confounding factors, such as patients selection criteria, soft tissue treatment and definition of complications were also incomplete.

Quality of the evidence
The overall quality of the included studies is poor. Most of them are nonrandomized observational studies with serious limitations. There was evidence of publication bias for primary and secondary bone union, with higher union rates in bigger studies. Overall sample size allows obtaining several statistically significant results. However, the level of evidence of these findings is low or very low due to the heterogeneity of the pooled data and the risk of bias caused by the studies' design.

Conclusion
This study states the effectiveness of autologous graft for bone defects. Overall union rate was 91 % while union rate after additional procedures raised to 98 % in published studies. Available clinical evidence does not show a direct relation between bone defect size and bone union rate when autologous bone graft techniques were applied. Therefore, bone defect size should not be the only factor used when choosing between vascularized or non vascularized bone graft. Finally, pooled analysis stated that in the presence of infection, vascularized graft has a lower risk of post-surgery infection. Well-designed randomized, controlled trials are needed to raise the low level of evidence for those conclusions.