Union, complication, reintervention and failure rates of surgical techniques for large diaphyseal defects: a systematic review and meta-analysis

To understand the potential and limitations of the different available surgical techniques used to treat large, long-bone diaphyseal defects by focusing on union, complication, re-intervention, and failure rates, summarizing the pros and cons of each technique. A literature search was performed on PubMed, Web of Science, and Cochrane databases up to March 16th, 2022; Inclusion criteria were clinical studies written in English, of any level of evidence, with more than five patients, describing the treatment of diaphyseal bone defects. The primary outcome was the analysis of results in terms of primary union, complication, reintervention, and failure rate of the four major groups of techniques: bone allograft and autograft, bone transport, vascularized and non-vascularized fibular graft, and endoprosthesis. The statistical analysis was carried out according to Neyeloff et al., and the Mantel–Haenszel method was used to provide pooled rates across the studies. The influence of the various techniques on union rates, complication rates, and reintervention rates was assessed by a z test on the pooled rates with their corresponding 95% CIs. Assessment of risk of bias and quality of evidence was based on Downs and Black’s “Checklist for Measuring Quality” and Rob 2.0 tool. Certainty of yielded evidence was evaluated with the GRADE system. Seventy-four articles were included on 1781 patients treated for the reconstruction of diaphyseal bone defects, 1496 cases in the inferior limb, and 285 in the upper limb, with trauma being the main cause of bone defect. The meta-analysis identified different outcomes in terms of results and risks. Primary union, complications, and reinterventions were 75%, 26% and 23% for bone allografts and autografts, 91%, 62% and 19% for the bone transport group, and 78%, 38% and 23% for fibular grafts; mean time to union was between 7.8 and 8.9 months in all these groups. Results varied according to the different aetiologies, endoprosthesis was the best solution for tumour, although with a 22% failure rate, while trauma presented a more composite outcome, with fibular grafts providing a faster time to union (6.9 months), while cancellous and cortical-cancellous grafts caused less complications, reinterventions, and failures. The literature about this topic has overall limited quality. However, important conclusions can be made: Many options are available to treat critical-size defects of the diaphysis, but no one appears to be an optimal solution in terms of a safe, satisfactory, and long-lasting outcome. Regardless of the bone defect cause, bone transport techniques showed a better primary union rate, but bone allograft and autograft had fewer complication, reintervention, and failure rates than the other techniques. The specific lesion aetiology represents a critical aspect influencing potential and limitations and therefore the choice of the most suitable technique to address the challenging large diaphyseal defects.

Data extraction. Two independent reviewers (PF, LS) screened all articles on the title and abstract and whether they met the inclusion criteria. After the first screening, the articles that met the inclusion criteria were evaluated on full-text eligibility and were excluded if they met one of the exclusion criteria (Fig. 1). In case of disagreement between the two reviewers (PF, LS) a third reviewer was consulted to reach a consensus (CC).
Data were independently extracted on a preconceived data extraction form using Excel (Microsoft). The following data were extracted: first author, journal, year of publication, level of evidence, population characteristics, cause of bone loss, surgical technique, graft characteristics, fixation method, surgery time, union rate, complications, reintervention, time to full weight-bearing, functional outcomes, and amputation. In case of missing data, an attempt to contact the corresponding author was made; in case of studies with data upon request, they were asked to the corresponding author. Delayed union was defined as the failure to reach bone union within 6 months after reconstruction, whereas non-union was defined as the failure to reach bone union at the time of the last follow-up after 6 months 12 . After independent data collection, the reviewers compared the extracted data.
Included articles were sorted in homogenous groups based on the surgical technique used. The interventions were classified into four main groups according to the treatment strategy: vascularized and non-vascularized fibular graft, bone autograft and allograft, bone transport, endoprosthesis. Comparable outcomes were also analysed between different groups of techniques, and their results were descriptively discussed. A meta-analysis was performed focusing on union, complication, re-intervention, and failure rates.

Assessment of risk of bias and quality of evidence. The Downs and Black's "Checklist for Measuring
Quality" 13 and the Cochrane Collaboration Risk of Bias (RoB) 2.0 tool was used to evaluate the risk of bias 14 . The first contains 27 'yes'-or-'no' questions across five sections; and provides a numeric score out of a scale of 32 points. The five sections include questions about the overall quality of the study (10 items), the ability to generalize the findings of the study (3 items), the study bias (7 items), the confounding and selection bias (6 items), and the power of the study (1 item). Rob 2.0 is designed into a fixed set of bias domains, focusing on different aspects of trial design, conduct, and reporting. Within each domain, a series of questions ask information about features of the trial relevant to the risk of bias. A proposed judgement about the risk of bias arising from each domain is generated by an algorithm, based on answers to these questions. The risk of bias can be judged as 'Low' , 'Some concerns' , or 'High' . The quality of evidence for all outcomes was graded using the Grading of Recommendations Assessment, Development, and Evaluation, which classifies the quality of evidence as high, moderate, low, or very low 15 . Evidence from RCT will start at high quality and be selected to be downgraded by 1 or 2 levels Statistical analysis. The statistical analysis was carried out according to Neyeloff et al. 16

using Microsoft
Excel. The Mantel-Haenszel method was used to provide pooled rates across the studies. A statistical test for heterogeneity was first conducted with the Cochran Q statistic and I 2 metric and was considered the presence of significant heterogeneity with I2 values ≥ 25%. When no heterogeneity was found with I2 < 25%, a fixed-effect model was used to estimate the pooled rates and 95% CIs. Otherwise, a random-effect model was applied, and an I 2 metric was evaluated for the random effect to check the correction of heterogeneity. The studies' rate confidence intervals were carried out using the continuity-corrected Wilson interval. The influence of the various techniques, as divided into the four groups, on union rates, complication rates, and reintervention rates was assessed by a z test on the pooled rates with their corresponding 95% CIs. Descriptive statistics were performed to describe the sociodemographic and injury-related characteristics of the patients included in the retrieved studies. Continuous variables were expressed as pooled means with their confidence intervals and standard deviation, using Excel (Microsoft).   Table 2. Gender was represented by 70.4% men and 29.6% women, and age presented a range of 2-86 years. The mean follow-up was 40.9 months (range 1 to 157 months). Results obtained for each treatment group are analysed in detail in the following paragraphs. Table 3 shows a summary of the results. Moreover, data are also reported quantifying the primary union, time to union, complication, reintervention, and failure rates of the different treatments based on the main aetiology subgroups trauma and tumour (see Tables 4 and 5 for details). A summary of complications is reported in Table 6.
Bone allograft and autograft. Bone allograft and autograft was used in 27 study arms 17-43 for a total of 564 patients (mean age 35.0 years): the inferior limb was involved in 78.2% of the cases, while the upper limb accounted for 21.8% of the cases. The defect was exclusively due to trauma in twelve studies, non-union in nine studies, tumour resection in five studies, and infection in one study. The average bone defect was 7.3 cm (range 0.5-27.8 cm), and the average follow-up was 33.3 months (range 6-180 months). A two-stage induced membrane technique (IMT) was used in seven studies, with massive autologous cancellous bone graft as a second step after the polymethacrylate spacer, while in nine studies cortical-cancellous iliac bone autograft was used, fixed either with external or internal fixation devices. In two studies autogenous cancellous bone graft was used alone, fixed with a plate and screws, while intercalary allograft with cancellous bone graft was used in two studies. Two studies reported the usage of titanium mesh cages with autologous bone graft, two studies the reimplantation of extracorporeal devitalized cancerous bone, one study reported the use of BMP-2, one study reported the application of multiple wrapped cancellous bone autograft methods, and another study reported a solution with an intramedullary nail and polymethacrylate spacer as a definitive treatment. Four studies did not report data on primary unions; in the others, the pooled union rate was 75% (C.I. 72%-78%). The average time to union was 7.8 months (range 1-27 months). Complications were reported 182 times, with a mean of 26% (C.I. 22%-30%) (details in Table 6). The reintervention rate was 23% on average (C.I. 19%-28%), and the failure rate was 8%     www.nature.com/scientificreports/ method with external fixation was used in 15 studies 35,36,[46][47][48][49][50]52,56,[59][60][61][62]65 , intramedullary nailing and/or external fixation was used in eight studies 43,45,51,[53][54][55]57,64 , a monolateral external fixation was used in one study 44 , and bone transport with a five-ring circular external fixator in one study 58 . One of the studies in the Ilizarov group was a case-control comparing bifocal versus trifocal fibular transfer, showing better primary union rate, fewer complications and reinterventions in the trifocal group. The primary union was obtained in 91% of the patients (C.I. 89%-93%); the mean external fixator time was 8.9 months (range 3-52 months). Complications were reported 425 times, with a pooled ratio of 62% (C.I. 59%-65%) but most of them were minor complications and did not require any invasive intervention (details in Table 6). The reintervention rate was 19% (C.I. 16%-22%), and the failure rate was 8% (C.I. 6%-10%). Outcomes of trauma treatment are detailed in Table 4.
A subgroup of six studies 45,46,50,51,53,54,57 focusing on the bone transport approach showed a primary union rate of 89.7% with a mean external fixator time of 6.6 months (range 4.2-9.3 months), a complication rate of 66.9%, a reintervention rate of 29.8% and a failure rate of 3.7%.
Vascular and non-vascular fibular graft. The fibular graft was the elective technique in 19 studies 37,42,43,[66][67][68][69][70][71][72][73][74][75][76][77][78][79][80][81] , for a total of 327 patients (mean age 22.2 years): the inferior limb was involved in 81.3% of the cases, while the upper limb accounted for 18.7% of the cases. The defect was exclusively due to tumor resection in nine studies, trauma in four studies, and non-union after previous treatment in three studies, while three studies reported mixed aetiology. The average defect was 13.4 cm (range 1-25 cm). The average follow-up was 62.8 months (range 6-276 months). In all except four studies, the graft was a vascularized fibular graft, in some cases used in combination with cortical-cancellous bone autograft or allograft. The primary union was achieved in 78% of the patients (C.I. 73%-82%), in an average time of 8.3 months (range 2-33 months). Complications were reported 130 times, with a complication rate of 38% (C.I. 33%-43%) (details in Table 6). The reintervention rate was 23% on average (C.I. 19%-28%), and the failure rate was 8% (C.I. 5%-12%). Outcomes of trauma and tumor treatment are detailed in Tables 4 and 5 1-306 months). A meta-analysis on the primary union rate was not performed due to the lack of data in the retrieved studies; however, it was possible to calculate the complication rate (mean 26%, C.I. 21%-32%), total complications 60, and the reintervention rate (mean 20%, C.I. 15%-28%). The meta-analysis on failure rate showed a mean of 22% (C.I. 18%-28%). 13 gives each study an excellent ranking for points ≥ 26, a good ranking for points between 20 and 25, a fair ranking for points between 15 and 19, and a poor ranking for a score ≤ 14 points. Accordingly, among the retrieved studies four studies were classified as poor, 50 studies as fair, and ten studies as good (Fig. 3). The main factors influencing the study quality was the inaccuracy of some studies in reporting data and results. The Rob 2.0 tool 14 reported that one study was to be considered at "low risk of bias", 53 studies with "some concerns for bias", and 20 studies at "high risk of bias" (Fig. 4). Based on the GRADE tool, the quality of evidence of all the four primary outcomes (primary union, complications, reintervention, and failures) was judged ranging from "low" to "very low".

Discussion
The main finding of this meta-analysis is that different treatments offer suitable results to address the complex diaphyseal bone defects, but each one showing specific indications, strengths, and critical aspects. Despite an overall high final union rate, relatively high complication and reintervention rates were retrieved, with important differences among the various treatments, which should be considered when choosing the proper surgical approach.
The correct approach for a LDD should be chosen considering each patient individually, weighing the pros and cons of each technique, aiming to achieve safe and reproducible outcomes with low reintervention rates. The results from this meta-analysis contribute to shed some light in this direction. In particular, the first finding brought to the attention by this systematic review is the inclusion of 74 studies for a total of 1781 patients; such large numbers underline the importance of diaphyseal bone defect treatment. A lot of research efforts have been put into this field, which also shows an evolution over time. A temporal trend emerged when investigating the various treatments for LDD (Fig. 2): during the 80 s and until the mid-90 s there were only a few studies, mainly focused earlier on bone allograft or autograft and, later, on vascularized or non-vascularized fibular grafts 17,41,66,67 . Then, from the end of the 90 s, more and more studies were progressively published, along with a growing interest in other approaches such as bone transport and endoprosthesis 54,57,[60][61][62][82][83][84] . Finally, in the last years, all treatments were increasingly addressed but with a focus above all on bone transport and, also due to the impulse of new augmentation procedures, with several studies on bone allograft or autografts [27][28][29]31,35,36,45,50,51,53,56,59,77,91 .
Traditionally, the surgical management of LDD involved bone grafting 92 , relying on the combination of mechanical stability and an osteoinductive substrate. In this meta-analysis, bone allografts or autografts have been used alone or within more advanced biological augmentation techniques. Among the retrieved studies, the mean primary union rate was 75%, an important finding both in terms of an overall good outcome, but also for the low heterogeneity of the findings, underlying the reliability and consistency of the results obtained by this www.nature.com/scientificreports/  www.nature.com/scientificreports/ approach  . Moreover, while a similar time to union and reintervention rates were found for other techniques, this approach presented one of the lowest complication and failure rates (Table 3). Interestingly, results obtained with bone allografts or autografts used alone [17][18][19][20]22,24,25,30,33 seemed to align with those of more complex combined procedures, apart from a major rate of complications, underlining the need for better-targeted studies to demonstrate the real potential of biological augmentation techniques for cancellous or cortical-cancellous bone grafting. Finally, in the early 2000s, the Masquelet technique started to raise interest, a two-stage reconstruction technique based on cancellous bone graft and induced membrane technique 93 . Since then, the induced membrane technique was increasingly used; five studies where it was reported as a single technique could be included in this meta-analysis 28,29,36,39,40 , showing overall results comparable to those of the other autograft or allograft techniques. The use of autologous fibular graft has been also well documented over time, either vascularized or nonvascularized. The vascularized fibula was introduced in 1975 by Taylor et al. 7 to maximize the healing potential and bone viability while taking advantage of the possibility of treating simultaneously soft tissue damage through combined tissue flaps. In this meta-analysis only four studies used exclusively non vascularized fibula, while in the majority vascularized fibula was used, either pedicled or free. Overall, the non-vascularized fibula showed less satisfactory results in terms of primary union, complications, and reintervention rate as compared to the vascularized fibula. The use of a fibular graft proved to be a demanding surgery 94 , with the meta-analysis underlying similar results in terms of time to union and need for reinterventions, but the lower primary union and higher failure and complication rates with respect to other approaches. Moreover, the majority of complications were severe and mainly represented by fractures, which often healed at the last follow up but caused long-term discomfort and healing time. In light of these limitations, Capanna et al. 95 combined allograft and intramedullary VFG to strengthen the construct. However, even with technique modifications and the latest developments, fibular grafts remain a challenging surgical approach 96,97 .
The bone transport technique with distraction osteogenesis with external fixation is another well-known treatment introduced in 1969 by Ilizarov for bone lesions 98 . Nowadays, it remains one of the most useful and versatile approaches to address critical-size defects 99 , with the overall higher primary union rate among the possible strategies. However, there are well-known downsides: it takes several months and a fully compliant patient to be completed, and there is a high risk of complications 100 . These aspects emerged clearly from this meta-analysis; the mean primary union rate was 91% but with a complication rate of 62% 35,36, . The large majority of complications were due to infection, especially superficial pin tract infections that often resolved without any further operative intervention 49,54,56,57,59 . Other relevant problems were ROM limitations, limb length discrepancy and deformity. To overcome these problems, in the last decades new bone transport techniques using intramedullary devices were introduced, minimizing external fixation time and joint contractures, thus increasing patient compliance and satisfaction, while achieving bone union faster [101][102][103] . Finally, a comparison between bifocal or trifocal bone transport was reported in one study 50 , finding the first to be quicker, more effective, and with a lower complication rate. Further RCTs are needed to confirm the potential and limitations of new bone transport techniques for the treatment of diaphyseal defects.
In the case of short life expectancy or other contraindications to biological reconstruction, intercalary endoprosthesis has been proposed for diaphyseal defects 104 . Endoprosthesis provides a valid solution when aiming for early weight-bearing, immediate stability, and fast recovery, but this comes at the price of high risks of mechanical complications such as aseptic loosening and periprosthetic fractures 33 . Therefore, the main indication is for elderly cancer patients whose healing capacity is poor and the rapid restoration of function is more important than the long term durability 11 . This was confirmed by the meta-analysis, the mean age of endoprosthesis patients was 52.7 years old, greater than that of the other techniques, and all of the patients had a diaphyseal tumour [82][83][84][85][86][88][89][90] . The complication rate was 26%, the reintervention rate 20%, and, most important, a 22% failure rate was retrieved, confirming the limits of endoprosthesis as a long-lasting solution 105 . Future improvements aiming to better prosthesis stability, such as changes in prosthesis design or better fixation methods, are needed to provide a better solution in terms of outcome and complications for these complex and fragile patients.
Different techniques are applied for the treatment of different types of patients and lesions, which could influence the results in terms of outcomes and complications. Among the four groups of techniques retrieved, the bone defect causes that led to the surgical intervention were different: bone allograft and autograft and bone transport were used mainly for traumatized patients (53.7% and 55.6%), almost all patients with a diaphyseal bone defect caused by infection were treated with bone transport techniques, while tumor resection was the indication of 73.1% of the fibular graft and 100% of the endoprosthesis ( Table 2).
In this light, a specific investigation was performed to define the results of the treatment options when addressing the same etiological target. The main aetiology groups are represented by tumor and trauma (Tables 4  and 5). Concerning tumors, it is interesting to underline the lower indications for some techniques: bone transport was not applied and when bone allograft and autograft, as well as fibular grafts, were used, they offered lower results than for other treatment indications. Furthermore, this sub-analysis revealed that, despite the previously mentioned limitations, endoprosthesis still presents the lower complication and reintervention rates with respect to the other options (Table 5). Finally, a specific analysis was focused on the trauma that was the most common cause of LDD. In these patients, similar results were achieved in terms of primary union rate, but marked differences were found for different treatment options (Table 4). Fibular grafts provided a faster time to union with respect to other treatment indications and showed to gain a two-to-three months advantage versus other solutions for patients affected by trauma (Table 4). On the other hand, high complication, reintervention, and failure rates were documented in these patients, which suffered fewer complications, reinterventions, and failures when treated with bone allografts and autografts (Table 4).
This systematic review and meta-analysis included numerous studies and a high number of patients. Nonetheless, there are limitations to point out. Among all, the risk of bias and the low level of evidence of the studies, as documented by the Downs and Black's Checklist for Measuring Quality, the Rob 2.0 tool and the GRADE system. www.nature.com/scientificreports/ Another limitation is the lack of prospective PROSPERO registration. Furthermore, the high heterogeneity of the included studies for defect aetiology, patient populations, and techniques, add to the literature weaknesses.
In light of all this, there is a clear need for standardized, properly designed comparative studies to search for the best clinical practice in each specific scenario. However, important indications could be drawn from the metaanalysis of the available literature. First of all, this meta-analysis documented that many options are available to treat LDD, but no one appears as an optimal solution in terms of safe, satisfactory, and long-lasting outcomes, which urges the development of better surgical strategies. Moreover, many aspects have to be taken into account when choosing the most suitable approach for LDD, as results in terms of primary union rate and time to union time, as well as risks in terms of complication, reintervention, and failure rates. The results of this meta-analysis underlined the potential and limitations of the different treatment options according to the different clinical scenarios, which could help the clinicians in understanding the advantages, disadvantages, and overall, the most suitable option when treating the challenging LDD.