Management of critical-sized bone defects in the treatment of fracture-related infection: a systematic review and pooled analysis

Purpose This systematic review determined the reported treatment strategies, their individual success rates, and other outcome parameters in the management of critical-sized bone defects in fracture-related infection (FRI) patients between 1990 and 2018. Methods A systematic literature search on treatment and outcome of critical-sized bone defects in FRI was performed. Treatment strategies identified were, autologous cancellous grafts, autologous cancellous grafts combined with local antibiotics, the induced membrane technique, vascularized grafts, Ilizarov bone transport, and bone transport combined with local antibiotics. Outcomes were bone healing and infection eradication after primary surgical protocol and recurrence of FRI and amputations at the end of study period. Results Fifty studies were included, describing 1530 patients, the tibia was affected in 82%. Mean age was 40 years (range 6–80), with predominantly male subjects (79%). Mean duration of infection was 17 months (range 1–624) and mean follow-up 51 months (range 6–126). After initial protocolized treatment, FRI was cured in 83% (95% CI 79–87) of all cases, increasing to 94% (95% CI 92–96) at the end of each individual study. Recurrence of infection was seen in 8% (95% CI 6–11) and amputation in 3% (95% CI 2–3). Final outcomes overlapped across treatment strategies. Conclusion Results should be interpreted with caution due to the retrospective and observational design of most studies, the lack of clear classification systems, incomplete data reports, potential underreporting of adverse outcomes, and heterogeneity in patient series. A consensus on classification, treatment protocols, and outcome is needed to improve reliability of future studies.


Introduction
Segmental bone loss after trauma remains a challenging problem for orthopedic trauma surgeons. When a bone defect exists combined with fracture-related infection (FRI), the chances of successful bone consolidation and clearance of infection are reduced.
A 'critical-sized' defect is a bone defect which is not expected to heal in the absence of a secondary (surgical) intervention. There is no agreed definition of what constitutes a critical defect in humans. Court-Brown defined it [1] as a defect involving 50% of the cortical diameter with a minimum length of 1 cm, and this was used in the Study to Prospectively evaluate Intramedullary Nails in Tibial fractures (SPRINT) [2]. A study by Sanders et al. [3] showed that, when using this definition, 47% of the bone defects healed without additional surgery, thus indicating that these are not always critical defects. When infection is present, it is much less likely that the fracture will heal [4,5], so in this analysis, The SPRINT definition for a critical-sized defect in FRI was accepted.
Over the past decades, critical-sized bone defects in FRI have been treated using different protocols. Techniques used (e.g., Ilizarov, Papineau, Masquelet, or RIA) all have different indications and success rates, resulting in a wide range of clinical outcomes. The aim of this systematic review and pooled analysis was to evaluate identified treatment strategies, their individual success rates and other outcome parameters regarding critical-sized bone defects used between 1990 and 2018.

Literature search strategy
A literature search was completed with the help of a biomedical information specialist on June 25, 2018, using Medline, Embase, Web of Science, Cochrane, and Google Scholar. The search strings are provided in Appendix 1. Studies that described treatment of FRI using autologous cancellous grafts, autologous cancellous grafts combined with local antibiotics, the induced membrane technique, vascularized grafts, bone transport, and bone transport combined with local antibiotics were included. Series needed to be greater than five patients, reported in English, and bone defects described as ≥ 1 cm. Studies that did not describe FRI patient treatment and publications reporting non-original data (e.g., reviews or meta-analyses), or those published before 1990, were excluded. Inclusion was agreed by two independent reviewers, HB an LWV, and consisted of two phases. During the first phase, title and abstract were screened for relevance and full text articles were obtained. When possible, full texts that were not available, were obtained by contacting the corresponding author once by email. The second phase consisted of reviewing the full text articles. Consensus was reached on all references. This study was performed using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) criteria [6].

Data extraction
Data from each included study were extracted by two authors independently (HB an LWV). Disagreements were discussed until agreement was reached. Data were collected in three areas.
1. General information of all studies (i.e., sample size, age, FRI, and location of FRI).
2. Data from surgical protocols (i.e., number of stages in surgical protocol, bone defect size, type of bone graft and type of fixation used).

Clinical outcomes (bony consolidation without infection
after the primary surgical protocol, bony consolidation without infection after the study period, recurrence of FRI, amputation of the affected limb, number of complications, revision surgery, time to bony union, and Length of Hospital Stay (LOHS).

Analysis
A quality assessment of all included studies was done according to the revised and validated versions of MINORS [7]. Results for the total population were pooled and presented separately for each of the six different treatment strategies. Medcalc (MedCalc Statistical Software version 17.9.7, MedCalc Software bvba, Ostend, Belgium; http:// www.medca lc.org; 2017) was used for pooling binominal data. Heterogeneity was quantified using the Cochran's Q test and I 2 statistic. When I 2 was < 40%, a fixed effects model was used, and a random effects model, when I 2 was ≥ 40%. Pooled estimates are reported with their 95% confidence intervals (CI). Publication bias was assessed from funnel plots for each clinical outcome and per treatment type separately. The majority of studies only provided a mean but not the standard deviation. Thus, a full meta-analysis for continuous data was not feasible. Continuous data were pooled by calculating the weighted mean using Microsoft Excel. Sample size of the individual studies was used as weighting factor. The pooled mean is reported with the range.

General population demographics
After selection, 43 studies  were included describing 50 patient series in the treatment of FRI with bone defects of ≥ 1 cm (Fig. 1). In these studies, 1530 patients with FRI were treated, with a mean bone defect of 6.6 cm (range 1.0-26.0), of which 1253 (82%) were localized at the tibia. The population had a mean age of 40 years (range 6-80), with 1176 (79%) male patients and a mean duration of infection of 17 months (range 1-624). Mean follow-up was 51 months (range 6-126). Study characteristics and quality assessment according to the MINORS score are given in appendix 2. The type of bone grafts used are depicted in Table 1. Nine series (18%) described treatment of bone defects by cancellous grafts alone and nine series (18%) by the use of a cancellous graft in combination with local antibiotics, either beads or an antibiotic rod. The induced membrane technique was described in eight (16%) series, and treatment with a 1 3 vascularized bone graft was used in four series (8%). Bone transport alone was used in 15 series (30%), and bone transport in combination with local antibiotic beads in 5 (10%). Table 2 summarizes the population of all treatment types and  Table 3 explains all different surgical stages.

Vascularized grafts
In four (8%) studies [16,21,41,46], a total of 322 (21%) patients were treated with a vascularized bone graft all using mixed, internal and external, fixation protocols. Two studies (50%) used a single-stage procedure, one study (25%) a two-stage procedure, and one study (25%) did not describe the number of stages. Their mean patient age was 43 years (range 6-69), with a mean defect size of 10.7 cm (range 1.0-21.0), and a mean follow-up of 75 months (range 6-86).

Clinical outcome of all studies
After initial protocolized treatment, FRI was cured and bone defects healed in 83% (95% CI 79-87) of all cases, increasing to 94% (95% CI 92-96) with further treatment, at the end of the total study period. Recurrence of infection was seen in 8% (95% CI 6-11), and amputation in 3% (95% CI 2-3) of all cases. Figure 2 shows the clinical outcome for all studies. Final outcomes overlapped across all different treatments. The use of the induced membrane or bone transport technique shows a trend of increased recurrence of infection when compared to the other techniques. Reconstruction with the induced membrane technique, a vascularized graft, or bone transport with local antibiotics tends to lead to a larger amputation rate than the bone transport and two other cancellous graft treatments alone. Funnel plots assessing publication bias for each clinical outcome are presented per treatment type in Appendix 3. The funnel plots of primary and total healing both show comparable asymmetry, possibly due the smaller studies having worse outcome. Publication bias could have occurred because of different study size populations in mostly retrospective series underreporting negative outcomes. This could also be the case for the funnel plots depicting recurrence of infection and amputation ratio. Half of all studies do not report recurrence of infection, possibly due to better surgical protocols, or underreporting of negative clinical outcome in mostly retrospective series. Forest plots depicting clinical outcome per treatment type are given in Appendix 4. Figure 3 shows the weighted means for all continuous outcomes. In 33 (66%) studies reporting time to union, bony union was achieved after a mean of 6.7 months. Patients were hospitalized for a mean of 1.7 months. A mean of 0.7 complications per patients were recorded, i.e. superficial or deep infection, hemorrhage, deformities, and non-union. Almost one-third of all patients required at least one reoperation. Overall time to union, LOHS, complication rate per patient, and surgical revisions per patient could not be compared between groups. The use of vascularized grafts tends to have a shorter time of union, number of complications and less necessary surgical revisions when compared to the other groups.

Discussion
Adequate debridement of non-viable bone remains one of the cornerstones of the management of FRI [51,52]. This debridement, along with the bone loss that occurs at the time of the initial trauma and due to the infectious process itself, frequently results in the presence of significant bone defects in patients with FRI. The optimal management of these bone defects remains controversial, leading to the numerous techniques described [53].
This systematic review revealed that the current most popular techniques resulted in comparable rates of healing and recurrence of infection. It is however unlikely that one reconstruction technique will be able to address all types of bone defects in FRI patients. Kadhim et al. [54] reviewed the use of bone transport or a vascularized bone graft for reconstruction of segmental bone defects, showing different success rates at different anatomical locations. Other recent reports have raised concern with regards to the efficacy of the induced membrane technique, for example, as a reconstruction technique of post-infective tibial shaft defects [30,55,56]. The need therefore is to determine what technique would be deemed optimal in a certain specific clinical scenario.
Due to several factors, it was not possible to draw firm conclusions from the data in this regard. Firstly, the heterogeneity of the patient populations made comparison of outcomes difficult. There was, for example, a difference in the mean size of the defects among the different treatment groups. Series describing the use of vascularized grafts or bone transport techniques treated larger bone defects than series using cancellous grafts. Also, in studies describing a two-stage reconstruction protocol, often more surgical revisions were performed than initially expected.
The choice of reconstruction technique depends on many factors. These include the host's physiological status and ability to participate in the rehabilitation program, the shape and location of the defect, duration of the defect (i.e., acute/ early or chronic/late-onset), quality of the surrounding soft tissue, bone quality, the presence of deformity, adjacent joint contracture/instability or limb length discrepancy, as well as the experience of the surgeon [57]. Unfortunately, the published series do not always report the details of these factors. It is often impossible to understand how patients Most studies provide a cohort of cases treated with a single technique, rather than a consecutive series of patients, selected because of the characteristics of the disease [58].
This makes firm conclusions on the efficacy of a technique in a patient population impossible [59].
Secondly, of all included series, 12 (24%) [20,23,32,33,35,37,40,41,43,46,47] describe FRI on multiple anatomical locations, and six (12%) [8,12,17,38,48] describe FRI of the femur, ankle or foot. Although these series are low in number of patients, data pooling still results in a distorted outcome. The current lack of a universally accepted classification system for both FRI and post-infective bone defects is evident in this systematic review. The need remains for the development of a pragmatic classification system that can be used, not only to guide treatment, but also to adequately characterize the patient cohort for research purposes [60][61][62]. Similarly, traditional classification systems for post-infective bone defects have failed to keep up with contemporary trends in reconstruction. The classification of post-infective tibial bone defects proposed by May and Jupiter, for example, does not cater for the induced membrane technique [63].
In addition to the lack of an applicable classification of post-infective bone defects, a working definition of defects is needed which cannot be expected to heal without intervention. While the difference between a stable cavitory (contained) defect and an unstable defect may be apparent, there may be a need to further define critical unstable defects to investigate the outcomes of treatment modalities in different types of critical bone defects. This definition, together with a pragmatic classification would allow the development of a practical algorithm for management of these infected defects. It may be foreseen that a 4 cm conical (partial) humeral defect with 20% cortical bone contact may be treated differently than a 4 cm segmental (complete) tibial defect, for example [64].
In addition to the lack of uniformity in terms of patient population, numerous other factors prohibit direct comparison of results from the studies included in this systematic review. Due to the retrospective and observational nature of most studies, treatment strategy selection was not always clearly reported and may not have been consistent. Furthermore, data were not always sufficiently complete to allow comparison of outcomes. For continuous outcomes, most studies failed to report the standard deviation along with the mean value. Underreporting of adverse effects is of particular concern, especially if risks associated with a specific treatment option in a specific clinical setting should be taken into account when choosing a reconstructive option. When looking at the asymmetrical form of the funnel plots created for all clinical outcomes, this could be the case, as is discussed above.
Mauffrey and Hak [55] have recently highlighted the lack of standardization in the management algorithms for tibial defect reconstruction, particularly in the setting of postinfective defects where treatment strategy selection is often based on surgeon preference rather than scientific evidence [65]. Similarly, Makridis et al. [65] found great heterogeneity of studies in the field of FRI, which made the development of evidenced-based management protocols difficult. The lack of comparative studies is also evident. While bone transport and the induced membrane technique remain two popular treatment options, there is no high-quality evidence indicating that one is superior to the other [66], or which patients are best suited to each technique. This systematic review confirmed that evidence to make any high-level recommendation with regard to the management of post-infective bone defects in FRI patients is currently insufficient.
It is unlikely that one single, high volume center is able to adequately study the problem of FRI. Even in such referral centers the number of patients treated in a reasonable time span is not sufficient to achieve enough statistical power to draw firm conclusions on current treatment standards and new therapeutic options. This implies that (international) collaboration is essential to be able to pool treatment results from individual hospitals into (prospective) clinical studies and subsequently into meaningful meta-analyses. Only then, can adequate progress in the treatment of FRI be made. To facilitate such a collaboration, uniform definitions for classification, diagnostic and treatment protocols and follow-up are required. The current meta-analysis clearly shows that the way in which data have been collected and reported in the past cannot inform best practice in the future.

Conclusions
This is the first extensive review of bone defect treatment protocols for FRI. Six individual treatment protocols for FRI treatment were identified, i.e., cancellous grafts with or without the use of local antibiotics, the induced membrane technique, the use of vascularized grafts, or the Ilizarov bone transport technique with or without the use of local antibiotics. All show comparable outcome. Overall published work showed a high success rate of 94%, a recurrence rate of 8%, and low amputation rate of only 3%. However, data did not allow a reliable comparison across treatments, or a recommendation on which treatment strategy is appropriate for any particular clinical scenario. The results should thus be interpreted with caution due to the retrospective design of most studies, the lack of clear classification systems, incomplete data reports, underreporting of adverse outcomes, and heterogeneity in patient series. Secondly, this review reveals the true scientific and clinical needs: uniform definitions on terms used, a consensus on classification, structured treatment protocols, and clear outcome parameters are needed to improve reliability of future multicenter studies. critically revised the manuscript, and approved the final version. EMMVL attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted. EMMVL is the guarantor of the paper and takes responsibility for the integrity of the work from inception to published article.
Funding There is no funding source.

Compliance with ethical standards
Conflict of interest The authors declare that they have no conflict of interest.
Ethical approval This article does not contain any studies with human participants or animals performed by any of the authors.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/.