Therapeutic Efficacy and Safety of Osteoinductive Factors and Cellular Therapies for Long Bone Fractures and Non-Unions: A Meta-Analysis and Systematic Review

Background: Long bone fractures display significant non-union rates, but the exact biological mechanisms implicated in this devastating complication remain unclear. The combination of osteogenetic and angiogenetic factors at the fracture site is an essential prerequisite for successful bone regeneration. The aim of this study is to investigate the results of the clinical implantation of growth factors for intraoperative enhancement of osteogenesis for the treatment of long bone fractures and non-unions. Methods: A systematic literature review search was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines in the PubMed and Web of Science databases from the date of inception of each database through to 10 January 2022. Specific inclusion and exclusion criteria were applied in order to identify relevant studies reporting on the treatment of upper and lower limb long bone non-unions treated with osteoinductive or cellular factors. Results: Overall, 18 studies met the inclusion criteria and examined the effectiveness of the application of Bone Morphogenetic Proteins-2 and -7 (BMPs), platelet rich plasma (PRP) and mesenchymal stem cells (MSCs). Despite the existence of limitations in the studies analysed (containing mixed groups of open and close fractures, different types of fractures, variability of treatment protocols, different selection criteria and follow-up periods amongst others), their overall effectiveness was found significantly increased in patients who received them compared with the controls (I2 = 60%, 95% CI = 1.59 [0.99–2.54], Z =1.93, p = 0.05). Conclusion: Administration of BMP-2 and -7, PRP and MSCs were considered effective and safe methods in fracture treatment, increasing bone consolidation, reducing time to repair and being linked to satisfactory postoperative functional scores.


Introduction
Long-bone fractures, including femoral, tibial and humeral, represent one of the most frequent types of non-fatal traumas worldwide [1]. Although criteria for conducting the 2 of 28 epidemiological surveys about musculoskeletal injuries vary from study to study, the overall prevalence of long-bone fractures was estimated at 406 per 100,000 people per annum [2], being higher for the adult population [3]. The frequency of humeral fractures in particular, which account for 0.5 to 3% of all fractures [4], is increasing with population aging, with significant socioeconomical effects [5].
Overall post-operative outcomes and complication rates are not well established for these injuries and depend on anatomic location, the severity of the accompanying soft-tissue injury, the patients' comorbidities and the fracture fixation technique [6]. Prospectively collected UK national data demonstrated that the rate of non-union after long bone fracture treatment was between 1.9 and 10%, and depended on the type of fracture and the age group [6,7]. The same study group also analysed the fractures of long bones that were treated between 2005 and 2010 in Scotland; it reported that the overall prevalence of non-union was 18.94 per 100,000 population per annum [8]. Similarly, in the USA, the annual incidence of fracture non-union was estimated at 100,000 cases [6,9]. Regarding the association between anatomical fracture location and compromised bone healing, it was reported that the overall non-union rate of tibial, femoral and humeral shaft fractures after intramedullary nailing was 4.6% [6,10], 8% [6,11] and 33% [4,6] respectively. Moreover, analysis of 1106 cases of tibial fractures treated with reamed intramedullary nailing showed that the non-union rate of open fractures for Gustilo-Anderson type I-II and type IIIB was 42.1% and 69.2%, respectively [6,12].
Strategies to enhance bone repair during fracture non-union treatment include autologous or synthetic bone and allogeneic or xenograft grafting, implantation of growth factors, progenitor cells and/or combination of graft materials. Fixation techniques that apply in the treatment of long bone fractures include intramedullary nailing (IMN), open reduction and internal fixation (ORIF) and external fixation. Ilizarov external fixation, used for the treatment of large bone defects, pseudarthrosis, limb deformities and lengthening, is a very popular intervention, and it is associated with early weight bearing and increased rates of beneficial functional outcomes. Moreover, it allows minor post-operative corrections of the limb axis. The effectiveness of the Ilizarov technique in the treatment of comminuted tibial fractures with large bone defects was confirmed by a recent study that reported beneficial clinical outcome scores for lower extremity function and relatively low rates of post-operative infections [13].
Current progress in the research of the molecular pathways involved in the osteoinductive and angiogenic processes of bone regeneration encouraged the clinical application of growth factors and cellular therapies in the treatment of non-union and comminuted fractures. Although many studies reported that the application of these factors was correlated with increased bone healing rates [6,[14][15][16][17], some others described low osteoinductive activity and a significant rate of complications such as post-operative infections [6,14]. Therefore, further research is needed to clarify the clinical potency of these therapies on patient recovery. The aim of this study is to compare the safety and effectiveness of the use of osteoinductive factors and cellular therapies for the treatment of close and open long bone fractures and non-unions.

Research Strategy
A systematic computer-based literature review search with predefined criteria was performed from the date of inception of each database up to 10 January 2022 according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [18] in the following repositories: PubMed (1947 to present) and Web of Science (1900 to present). The research methodology used a combination of the following terms: "long bone fractures [All Fields]", "Osteoinduction [All Fields]", "Bone Morphogenetic Proteins, BMPs [All Fields]", "Femur [All fields]", "Tibia [All fields]", "Humerus [All fields]", "Platelet Rich Plasma, PRP [All Fields]" and "Mesenchymal stem cells, MSCs [All Fields]". The electronic literature search was conducted independently by two authors (E.P., A.K.) and an experienced librarian. Moreover, the above two authors (E.P, A.K) independently screened the titles and abstracts to identify relevant studies of clinical outcomes and complications after the intraoperative application of angiogenetic growth factors for the treatment of long bone fractures. If there was a disagreement between them, the final decision was made by the senior author (P.V.G.).

Inclusion Criteria and Study Selection
Studies that analysed the clinical outcome in patients after the intraoperative application of growth factors and cellular therapy for the treatment of long bone fractures and non-unions were identified. Only full-text articles were eligible for inclusion. Additional inclusion criteria included: (a) studies written in English, (b) studies concerning the application of osteoinductive molecules in human subjects, and (c) data on the outcome clearly given to each patient.
Published studies written in a language other than English were excluded. Studies without obtainable data or insufficient details about the type of intervention and the clinical outcomes, case reports, reviews, letters to the editor, conference abstracts, technical notes, and expert opinions were excluded. Research based only on in vitro or in vivo animal model results was also excluded.

Data Extraction
Two reviewers (E.P. and P.V.G.) examined all identified surveys, extracting data using a predetermined form. All data of each study were assembled in a Microsoft Excel spreadsheet, classified by orthopaedic intervention and type of osteoinductive factor. Characteristics extracted from clinical studies included the first author, the publication year, study design (cohort or randomised control trial), enrolled sample number in both control and treatment groups, patient demographics, anatomical site of interest, type of fracture (closed or open), orthopaedic procedure, outcomes regarding the frequency of non-union development, type and rate of detected complications, type of growth factor and or cellular therapy used and length of follow-up period. As the included studies did not report in detail the mean time between the primary intervention and reoperation and since the primary goal of our study was to compare the prevalence of the reoperations between the groups, this factor was not analysed due to the lack of accurate data. Fracture healing was defined clinically as the absence of pain on loading and radiologically as the presence of bridging callus formation in three out of four cortices on plain X-rays. The presence of duplicate studies was examined using the Endnote software (Clarivate Analytics, Philadelphia, PA, USA). All data were collected, summarised and analysed by two independent authors, A.K. and A.C.H.

Quality Assessment
The methodology of each study was assessed independently by the two senior authors (E.P. and P.V.G.) using the Newcastle-Ottawa quality assessment scale [19]. Included studies were graded in a three-category scale. Studies displaying a total score of 0-3, 4-6 and 7-9 were classified as poor, fair or good quality, respectively. A modified Jadad scale for clinical trials [20] was also used to evaluate the quality of included trials. A Jadad score greater than 4 was considered to be of high quality.

Statistical Analysis
Statistical analysis was performed using the Review Manager (RevMan) Version 5.4 (Nordic Cochrane Centre, Cochrane Collaboration, Copenhagen, Sweden). Moreover, MedCalc Meta-analysis Statistical software, v. 17.2 (MedCalc Software's, Ostend, Belgium) was used to produce Egger's test in order to test funnel plot asymmetry. The incidence of long-bone healing after the application of osteoinductive factors as well as the odd ratios (ORs) and the associated 95% Confidence Intervals (95% CI) were calculated. Heterogeneity between the trials was calculated by using Cochrane Q and the inconsistency (I 2 )-test.
Values greater than 25%, 50% and 75% were considered as low, moderate and significantly heterogeneous, respectively. Therefore, a random effect model was used to calculate pooled ORs in the case of moderate and significant heterogeneity and the fixed effect model was used in the studies found with low heterogeneity. This was undertaken because, in the sensitivity analysis, the presentation of both models provides comprehensive evaluation of how differences in datasets affected the observed outcomes. Egger's test and Forest plots were used to examine the risk of publication bias. The level of statistical significance was set at p ≤ 0.05.

Search Results
The literature search and cross-referencing resulted in 4162 references. After the initial evaluation of the studies based on the abstract and title, 1534 publications were included. Further analysis of the remaining papers resulted in the exclusion of 191 studies, as they followed an in vitro methodology only. In addition, 141 studies were excluded because they referred to animal models. Based on the inclusion criteria, 528 studies were excluded after reading the full article, while 861 and 126 studies were excluded because they were review articles or were written in languages other than English, respectively ( Figure 1). Finally, 18 studies [21][22][23][24][25][26][27][28][29][30][31][32][33][34][35][36][37][38] published between 1996 and 2020 met our inclusion criteria. The data from each study are summarised in Tables 1 and 2. Specifically, the studies included for meta-analysis are displayed in Table 1, while the studies selected for qualitative evaluation are shown in Table 2. The degree of agreement among the reviewers who examined the scientific quality of the included studies was strong, as shown in Tables 3 and 4. odd ratios (ORs) and the associated 95% Confidence Intervals (95% CI) were calcul Heterogeneity between the trials was calculated by using Cochrane Q and the in sistency (I 2 )-test. Values greater than 25%, 50% and 75% were considered as low, mo ate and significantly heterogeneous, respectively. Therefore, a random effect model used to calculate pooled ORs in the case of moderate and significant heterogeneity the fixed effect model was used in the studies found with low heterogeneity. This undertaken because, in the sensitivity analysis, the presentation of both models prov comprehensive evaluation of how differences in datasets affected the observed outco Egger's test and Forest plots were used to examine the risk of publication bias. The of statistical significance was set at p ≤ 0.05.

Search Results
The literature search and cross-referencing resulted in 4162 references. After the tial evaluation of the studies based on the abstract and title, 1534 publications wer cluded. Further analysis of the remaining papers resulted in the exclusion of 191 stu as they followed an in vitro methodology only. In addition, 141 studies were exclu because they referred to animal models. Based on the inclusion criteria, 528 studies excluded after reading the full article, while 861 and 126 studies were excluded bec they were review articles or were written in languages other than English, respect ( Figure 1). Finally, 18 studies [21][22][23][24][25][26][27][28][29][30][31][32][33][34][35][36][37][38] published between 1996 and 2020 met our inclu criteria. The data from each study are summarised in Tables 1 and 2. Specifically, the s ies included for meta-analysis are displayed in Table 1, while the studies selected for q itative evaluation are shown in Table 2. The degree of agreement among the revie who examined the scientific quality of the included studies was strong, as shown in Ta 3 and 4.     Healing of non-union in 26     According to the Newcastle-Ottawa scale and the modified Jadad score, all included trials were considered of high quality and were therefore deemed to be at a low risk of bias (Tables 3 and 4). Moreover, after the evaluation of the funnel plot, all studies were found to lie within a 95% CI as represented by the inverted funnel, suggesting the absence of publication bias (Figure 2). According to the Newcastle-Ottawa scale and the modified Jadad score, all included trials were considered of high quality and were therefore deemed to be at a low risk of bias (Tables 3 and 4). Moreover, after the evaluation of the funnel plot, all studies were found to lie within a 95% CI as represented by the inverted funnel, suggesting the absence of publication bias (Figure 2).
Three studies analysed the efficacy of MSCs for sufficient osseous healing either in open tibial injuries associated with increased size of bone defects [21], in severe recalcitrant non-unions and pseudoarthrosis [24] as well as in previous infected tibial non-union fractures [25].
Similarly, five studies [22,23,26,28,31] assessed the effectiveness of PRP in accelerating the process of osseous healing. One study evaluated the activity of PRP in pseudoarthrosis healing [22]. A study compared the effectiveness of PRP and Hyperbaric Oxygen Therapy (HOT) for the treatment of aseptic tibial non-union using Ilizarov external fixation [23]. Due to the fact that a Cochrane systematic review did not reveal any clinical evidence supporting the efficacy of HOT application on the treatment of long bone fractures and non-unions [41], the HOT group was used as a control group in our analysis. Moreover, the correlation between fixation technique and PRP application was evaluated in three more studies [26,28,31]. One of them assessed the therapeutic potential of PRP in long bone oligotrophic non-unions treated using intramedullary nailing and whether exchange of implants was not reasonable. Likewise, the healing effect of PRP was interpreted in two further studies. The first analysed the application of PRP in delayed union of diaphyseal humeral fractures [28], while the second evaluated the role of PRP in long bone non-union fractures using IMN or ORIF fixation and autologous bone grafting.

Overall Effectiveness
The frequency of fracture healing after the application of osteoinductive factors in each of the included studies is shown in Figure 2.
The overall effectiveness was found significantly increased in patients who received them compared to controls (I 2 = 60%, 95% CI = 1.59 [0.99-2.54], Z =1.93, p = 0.05) (Figure 3). Regarding the healing rate at the fracture site, the patients who received PRP treatment for delayed union of diaphyseal humeral fractures as well as for pseudoarthrosis and oligotrophic non-unions of long bone fractures displayed no significant difference compared to the control groups (I 2 = 4%, 95% CI = 1.04 [0.97-1.12], Z = 1.13, p = 0.26). Moreover, the potential for bone healing by PRP administration was not affected by the preferred surgical technique as these patients were treated either with ORIF, IMN or an Ilizarov external fixator with or without the addition of autologous bone grafts ( Figure 4A). Similarly, we have observed that the application of MSCs or BMP-7 was not statistically associated with an improved incidence of efficient healing outcome between the treated and non-treated cases with long bone fractures (I 2 = 72%, 95% CI = 3. 23 Figure 4B). Likewise, the application of BMP-7 in aseptic diaphyseal ulnar and/or radial shaft nonunions treated with compression plating and autologous bone grafting as well as in tibial non-unios after IMN was not found therapeutic ( Figure 4C).
On the contrary, our meta-analysis showed that the application of BMP-2 (I 2 = 47%, 95% CI = 1.62 [1.21-2.18], Z = 3.24, p = 0.001) in patients with open tibial shaft fractures treated with reamed and unreamed IMN or an external fixator was correlated with significantly increased rates of sufficient bone healing compared to the control patients. Overall, the addition of BMP-2 revealed a great enhancement of the healing process when surgical options such as ORIF, IMN or external fixation were employed in femur, tibia and humerus non-united open or closed fractures ( Figure 4D).
It is worth mentioning that interventions at the fractured non-united site require the most appropriate and individualised preoperative planning, which combines the best surgical fixation technique with the application of the most suitable bone stimulating factor. Overall, the present meta-analysis demonstrates different effectiveness on the healing potential when growth factors such as BMPs, PRPs and osteogenic factors (MSCs) are applied on long bone non-unions or open fractures. For instance, the post-surgical outcomes after the application of growth factors in patients suffering from humeral, radial, ulnar, femoral and tibial non-unions were not found significantly improved compared to those whose surgical treatment did not include the above growth factors (I 2 = 17%, 95% CI = 0.88

Persistent Non-Unions and Re-Fractures after the Application of Osteoinductive Growth Factors
The present meta-analysis further assessed the rates of persistent long bone non-unions and re-fractures at non sufficiently healed sites after the administration or not of growth factors, such as BMP, PRP and osteogenetic factor (MSC) in patients suffering from long bone non-unions. Hence, PRP-treated patients suffering non-united long bone fractures were not experiencing higher rates of persistent non-unions or re-fractures compared to their counterparts (I 2 = 0%, 95% CI = 0.33 [0.14-0.76], Z = 2.62, p = 0.009), as shown in Figure 7A. On the contrary, the application of autologous MSCs in open tibial fractures (with gaps less than 10 mm) Gustilo-Anderson Type II or III as well as in closed non-united long bone fractures was correlated with a mild increase (I 2 = 72%, 95% CI = 0.31 [0.05-1.92], Z = 1.26, p = 0.21) of the risk for persistent non-unions and re-fractures, as shown in Figure 7B.  We have further considered the correlation between BMP application and persistent non-unions and re-fracture rates in patients suffering from non-unions after ORIF, IMN and external fixation treatment of closed and open long bone fractures. According to our meta-analysis, the administration of BMP-2 significantly (I 2 = 47%, 95% CI = 0.62, Z = 3.24, p= 0.001) prevents persistent non-unions and re-fractures in treated patients compared to non-treated ones ( Figure 7C). However, the application of BMP-7 at the fracture site in patients with long bone non-unions did not reveal any statistical difference (I 2 = 8%, 95% CI = 1.41 [0.74-2.69], Z = 1.03, p= 0.30) in the healing rate of persistent non-unions and re-fractures compared to their non-treated counterparts, as shown in Figure 7D.
In addition, the incidence of infections after application of bone stimulating factors was not significantly increased in patients who received PRP (I 2 = 0%, 95% CI = 2.

Discussion
It is well established that efficient bone healing develops through fundamental overlapping physiologic stages: haematoma formation, inflammatory reaction, chondrogenesis and angiogenesis, osteogenesis and remodelling [39][40][41][42][43]. Therefore, the combination of osteogenesis and angiogenesis is essential for bone regeneration-reported as the "coupling process" [43]. The present study compared the growth factors currently available in clinical practice and exhibiting a critical role in the induction of "osteogenesis and angiogenesis coupling". Although certain systematic reviews have already studied the involvement of BMPs, MSC and PRP in upper and lower extremity long bone fractures [44][45][46][47] BMPs are multifunctional cytokines belonging to the transforming growth factor-β (TGF-β) superfamily and their involvement in the osteogenesis process is well described [48][49][50][51]. In vitro and in vivo findings reported that BMP-2 promoted the angiogenetic effects in endothelial cells and induced the activation of circulating endothelial progenitor cells that possessed osteogenic and angiogenetic actions [49]. The present metaanalysis showed that the application of BMP-2 and BMP-7 at the fracture site was correlated with reduced fracture healing time, increased consolidation rates, and it was not accompanied by significant joint restriction or pain during motion. More specifically, according to a study by Govender et al., in patients with open tibial fractures administration of BMP-2 was accompanied by significant decrease of surgical failure risk, elimination of invasive interventions, such as bone grafting, a decrease in infections and secondary interventions rates compared to the control group [36,52]. This was also supported not only by the results of clinical cohorts ( Table 2), but also by recent research studies [53,54]. Indeed, Kostiv et al., using immunohistological methods, reported that fracture non-union was the result of overproduction of cytotoxic and proapoptotic factors in chronic inflammation and dysfunction of BMP-2 expression [53]. Additionally, the application of BMP-2 composites materials during fixation of displaced femoral neck fractures with cannulated screws provided fewer complications, such as avascular necrosis and nonunion [49]. Concurring with the literature [51][52][53][54][55][56][57][58][59], our meta-analysis also confirmed the results of previous studies ( Table 2) which supported the osteoinductive features of BMPs. However, our analysis showed that the increased healing rate, reduced repair time and post-operative infection rates were more prominent after the application of rhBMP-2 compared to rh-BMP-7 (Table 2). It is worth mentioning that consolidation rates after administration of rhBMP-2 appeared statistically higher compared to rhBMP-7 in two studies, the first by Haubruck et al. [50] and the second by Conway et al. [51]. Similarly, the systematic review of Sandler et al. confirmed that the application of BMP-2 in the treatment of upper extremity non-unions cases led to union in 117 days while in those treated with BMP-7 radiographic union was observed after 196 days [45].
Although adverse effects of BMP use were observed, such as heterotopic bone formation [55], synostosis and post-traumatic osteoarthritis [56], our meta-analysis showed that complication rates were low. This was in agreement with the results of Boraiah et al., who detected a limited number of heterotopic ossifications after rhBMP-2 application [52]. Additionally, no allergic or immunological reactions, severe deep infections, malignant transformation and persistent nerve palsy were observed. Our results were also consistent with the findings of several studies that reported limited side effects after the application of BMPs for the treatment of long bone fractures (Table 2). Indeed, the cohort studies of Kanakaris et al. [58,59], Giannoudis et al. [60] and Dimitriou et al. [61] that applied the combination of rhBMPs with autologous bone graft in a large sample for the treatment of persistent pseudarthrosis did not detect any infectious or allergic reactions.
Similarly, the risk for post operative hardware failure in patients receiving implant engraftment for their long bone fractures was overall significantly decreased after the application of BMPs. The above result was also supported by a study by Fuchs et al., which reported implant revision in only 3 out of 72 patients after the application of rhBMP-2 for the treatment of long bone non-union [62]. This could be explained by the fact that long bone fractures treated without BMP application were more prone to hardware failure due to either delayed healing or lower bone healing rates in general compared to BMP groups. It is particularly evident that the biomechanical stability of the construct is negatively affected in non-unions and is a major concern for future broken implants [63].
According to Caterini et al. [56], the administration of BMPs increased healing in nonunions of long bones either in one or two stage procedures, independently from the fixation technique (ORIF, IMN, external fixation, lag screws). However, our meta-analysis revealed that applied growth factors display a higher healing potential in fracture non-unions where IMN and external fixation (Ilizarov technique) are preferred, compared to ORIF.
Mesenchymal stem cells actively participate in angiogenesis and osteogenesis coupling through direct differentiation, cell contact interaction with endothelial lineage, and via releasing pro-angiogenic factors [64]. Furthermore, human adipose derived stem cells have an increased capacity to proliferate and differentiate into osteoblastic cells without the presence of growth factors [64]. These findings were also associated with the absence of genetic alterations, providing a safe method for clinical application [65]. Many studies (Table 2) demonstrated that MSC application at the non-union site was associated with increased fracture consolidation rates and with remarkably reduced fracture union times and hospitalisation periods [21,24,25]. The above findings were also associated with a high quality of health index [65]. Although the international literature reports that MSCs suppressed lymphocyte reaction and induced immunosuppressive properties [61], our analysis demonstrated that adverse effects after MSC application were extremely uncommon [21,24,25]. Despite the fact that many studies reported increased healing outcomes after the administration of MSCs in patients with atrophic and hypertrophic non-unions, in critical-size bone defects as well as in depressed tibial plateau fractures [66][67][68][69][70], our metaanalysis did not reveal statistical differences in the healing process or in the prevention of refractures after MSC application compared to the control group. Contrariwise, our analysis confirmed the safety of MSCs application as their use was not associated with allergic reactions, deep infections, ectopic neo-formations and neoplastic transformations [71].
Similarly, the application of PRP has also demonstrated an important role during the fracture repair process, due to its ability to induce complex inflammatory responses at the bone defect site, not only by activating angiogenesis but also by providing growth factors, such as VEGF, PDGF, TGF-β and IGF [71]. Although our meta-analysis did not reveal significant differences in the healing rate of patients with long bone fractures that received PRP treatment compared to the control group, the examination of the included studies showed that PRP use was accompanied by increased success rates for non-union treatment, decreased fracture healing periods and very good overall functional outcomes. Our findings are in line with the results of a recent meta-analysis that detected higher healing rates and a shorter consolidation duration accompanied by significant pain relief after the treatment with PRP for long bone delayed union and nonunion [45]. This was also supported by the study of Bielecki et al., where PRP use resulted in complete healing in all patients with delayed union fractures [72]. Moreover, no complications such as damage to blood vessels, haematomas, delayed wound healing, permanent nerve damage and post-operative infections were observed. Studies (Table 2) also demonstrated that PRP is a useful growth factor in non-unions of long bones, boosting the healing process without reported complications, such as allergic reactions, deep infections, ectopic neo-formations, neoplastic transformations or re-fractures, in line with our meta-analysis [73][74][75][76]. However, Calori et al. reported a superior osteoinductive activity of BMP-2 compared with PRP; this was consistent with our results [77].
An interesting finding of our analysis was that the efficacy of growth factors was statistically correlated with the fracture anatomical location. Indeed, the bone healing process was more prominent in tibial than in femoral and upper limb fractures. Our results were in line with a recent study that reported increased achievement of healing after BMP-2 in tibial fractures compared to humeral fractures [62]. A possible explanation to this finding could be that the relationship between tibial bone structure and axial mechanical loading may provide an exceptional biomechanical environment for growth factor activity. Mechanical loading is a significant factor of bone remodelling and the absence of mechanical signals was accompanied by increased bone resorption and reduced bone formation [77][78][79][80]. Furthermore, mechanical strain increased the secretion of chemokines and the recruitment of MSCs to bone surface, promoting bone formation [81]. Recent findings have also detected links between induction of endosteal progenitors and mechanical loading-induced growth factors released from osteoblastic cells [80]. As tibial bones receive increased axial loading [81], we can assume that these benefit the most from application of osteogenic factors after fracture.

Study Limitations
The study limitations include a language bias, as only studies written in English were reviewed and thus some studies were not included in our analysis. Another limitation was that the examined studies contained mixed groups of open and closed fractures and information about the number and outcome of closed fractures was based on data concerning only PRP application. The use of a HOT group as control group in the included study by Rollo et al. [23] could be another possible limitation. Further limitations were the variability of treatment protocols, the different selection criteria or follow-up periods. Additional limitations to consider were the differences in methodological approaches between the studies, the conditions under which these were conducted, other confounding factors that were not taken into consideration or the increased risk of bias of the selected studies, especially of those that were not evaluated. Additionally, the assessment of callus formation was based on clinical signs and plain radiographs that could be associated with uncertain healing outcome as CT-scan is the gold standard for the diagnosis of the healing process [59]. We must also draw attention to the fact that studies including thoroughly negative assessments about intraoperative growth factor application may face difficulties in being published by peer-reviewed journals.

Conclusions
Despite the limitations of our study, our results tend to support the efficacy of osteoinductive factors, such as BMP-2, BMP-7, PRP and osteogenic factors (MSCs) in the long-bone healing process. The intraoperative application of these factors was considered safe and with an effective intervention that enhanced angiogenesis and was linked to augmented fracture healing, reduced repair time and very good functional scores. Therefore, our meta-analysis suggests that the combination of a precise fixation technique with application of osteoinductive/osteogenetic factors supplies an optimum environment to support the healing microenvironment at the fracture site, with a positive impact on clinical outcomes.
Author Contributions: A.K., conceptualization, methodology, writing-original draft preparation; A.C.H., writing-original draft preparation; E.S.V., investigation, data curation, review and editing; D.J.P., writing-original draft preparation; P.V.G., supervision-review and editing; E.C.P., supervision-review and editing. All authors have read and agreed to the published version of the manuscript.