Nerve reconstruction with muscle-in-vein conduits VS autologous nerve grafts - a systematic review and meta-analysis of preclinical and clinical studies


 The gold-standard method for reconstruction of segmental nerve defects, the autologous nerve graft, has several drawbacks in terms of tissue availability and donor site morbidity. Therefore, feasible alternatives to autologous nerve grafts are sought. Muscle-in-vein conduits have been proposed as an alternative to autologous nerve grafts almost three decades ago, given the abundance of both tissues throughout the body. Based on the anti-inflammatory effects of veins and the proregenerative environment established by muscle tissue, this approach has been studied in various preclinical and some clinical trials. There is still no comprehensive systematic summary to conclude efficacy and feasibility of muscle-in-vein conduits for reconstruction of segmental nerve defects. Given this lack of a conclusive summary, we performed a meta-analysis to evaluate the potential of muscle-in-vein conduits. This work’s main findings are profound discrepancies regarding the results following nerve repair by means of muscle-in-vein conduits in a preclinical or clinical setting. We identified differences in study methodology, inter-species neurobiology and the limited number of clinical studies to be the main reasons for the still inconclusive results. In conclusion, we advise for large animal studies to elucidate the feasibility of muscle-in-vein conduits for repair of segmental defects of critical size in mixed nerves.


Introduction
Peripheral nerve injuries, which affect up to 60 out of 1.000.000 people in the US annually are frequently associated with devastating live-long sequelae for the affected patients and a high socioeconomic burden 1,2 . Neurotmesis, e.g. discontinuity of the entire nerve, always requires surgical treatment to reconstruct the severed nerve's continuity 3 . In cases of segmental nerve damage in which tension-free neurorrhaphy is impossible, autologous nerve grafts (ANGs) are considered the gold-standard treatment 4 . However, their use is restricted due to limited availability throughout the body and speci c requirements regarding length, diameter, and vascularization. Additionally, harvesting of a donor nerve results in loss of motor and/or sensory function which must be balanced against the expectable gain of function by means of the reconstructed recipient nerve 5 . Allogeneic nerve transplants can only be used in decellularized form due to the imminent graft-versus-host reaction and are also cost-intensive 6 .
Scientists aiming to develop an adequate alternative for autologous nerve grafts by means of tissue-and bio-engineering have devised and tested a multitude of biologic and synthetic materials to create an ideal conduit to bridge peripheral nerve defects 7 . Use of veins as biological nerve guidance conduits was pioneered by Wrede at the beginning of the 20th century 8  Although good results were reported for bridging of short defects, the vein's tendency to collapse over distances exceeding 1-2 cm 10 demanded a further re nement of the technique. In 1993, Brunelli and colleagues pioneered use of a muscle-in-vein conduit (MVC), combining the vein wall's shielding and antiin ammatory capabilities with the proregenerative effects of the muscle tissue 11 . MVC are prepared by harvesting a segment of an autologous vein and some muscle bers which are then pulled within the vein using ne forceps or a needle holder, yielding an autologous nerve guidance conduit. The MVC is then interposed between the severed nerve stumps, taking care to pull them 1-2 millimeters inside the vein to guarantee adequate entubulization 12 . The muscle bers' basal membranes act as natural guidance channels for regenerating axons while the muscle also prevents collapsing of the vein 13 . Following promising initial results obtained by Brunelli et al, the new reconstructive approach was further evaluated both in preclinical as well as clinical studies with increasing interest in the late 1990s and beginning of the 21st century 10,14,15 . Over the past three decades, multiple authors evaluated use of MVCs to reconstruct peripheral nerve defects in both animals and human patients, reporting divergent results.
While some authors conducting clinical studies did report non-inferiority of MVCs compared to ANGs 12,16 this has been opposed by reports of strong inferiority of MVCs in rodent models of segmental peripheral nerve injury 17,18 . To the best of our knowledge, one systematic meta-analysis which compared the effects of MVCs on the sensory recovery after digital nerve reconstruction to other reconstructive approaches has been published so far 19 . The outcomes were equivalent among all included reconstructive methods. However, this work was published in 2013 and since then a multitude of preclinical and some clinical studies regarding the use of MVCs has been conducted and published.
Given this increment both in reported patient outcomes and results from animal studies, a comprehensive analysis of both sets of data is still pending yet. In the light of the high need for alternatives to reconstruct peripheral nerves by means of grafting material, a comprehensive evaluation of the published data regarding MVCs will provide important insight into the actual feasibility and potential of this approach. This study therefore presents the rst systematic review and meta-analysis of studies evaluating MVCs to reconstruct peripheral nerves both in rats as well in human patients. It is the aim of this work to give an overview on the divergent outcomes after nerve repair by this technique and to discuss the potential mechanisms underlying these results. We aim to provide preclinical scientist and clinicians with a broad knowledge base regarding the published data on the use of MVCs for peripheral nerve repair.

Results
Results of the systematic literature research 10 studies, 8 preclinical and 2 clinical, met the inclusion criteria and were consecutively included in the quantitative synthesis. A total of 451 studies were identi ed, of which 183 remained for title and abstract screening after removal of duplicates. In accordance with the inclusion criteria (Table 1), 110 studies were excluded after title screening, whereas 40 studies were excluded after abstract screening. Therefore, 33 studies remained to be assessed for eligibility via full text screening. Two additional studies could be identi ed via reference checking. Out of these 35 studies, 25 studies were excluded due to various reasons. The detailed selection process is illustrated in Fig. 1. Eight preclinical studies were available in English and used either the sciatic nerve (n = 6) or median nerve (n = 2) injury model. A total of 245 animals was investigated (median nerve: n = 88, sciatic nerve: n = 157).
The smallest experimental groups consisted of 3 animals 11,20 while the largest comprised 14 animals 21 . The mean nerve defect size was 13 mm while the shortest and longest experimental nerve defect measured 7 mm 18 and 20 mm 11,22 respectively. The mean observation period was 20 weeks or 143 days, respectively. While several authors chose to observe the animals for 12 weeks, the longest observational period was 10 months 21 .

Clinical studies
The two clinical studies included were either conducted in patients with either median or ulnar nerve 16 or digital nerve injuries 12 in a total of 49 patients. The defect lengths ranged from 10 mm to 60 mm, which were either reconstructed with ANGs ( n = 26) or MVCs (n = 29). While the ANGs were harvested from the medial antebrachial cutaneous nerve 12 or sural nerve 16 , the origins of the MVCs were not further speci ed. In the one study the patients' age ranged from 11 to 72 years and patients were observed between 12 and 58 months 12 . The other study's authors reported a mean age of around 27 years, but the observation period was not speci ed 16 . Both studies assessed static two-point discrimination as common outcome measure, while dynamic two-point discrimination, Semmes-Weinstein mono laments, and muscle power according to the Medical Research Council Scale were also evaluated in either the one or the other study . Although the authors did not report any statistical analysis of this data, the published data sheet indicated no marked differences between the ANG and MVC group 20 .

Median nerve model
In the three studies evaluating MVCs versus ANGs in the rat median nerve model, motor function recovery was assessed in a total of 108 rodents by means of the grasping test. One study 18 with 16 rats featured the staircase test 23 as an additional assessment tool. Regarding the grasping test, in one study rats treated with MVCs recovered signi cantly (p < 0.05) less grasping strength than rats treated with ANGs at six months postoperatively 21 , while another study reported no signi cant differences between the two groups 24 . Stossel et al. did not report absolute grasping strength in grams but made use of a trinary scoring system to evaluate the rats' grasping ability. While they found no signi cant differences between the two groups at postoperative week (WPO) 4 and WPO12, signi cantly (p < 0.05) more animals treated with ANGs had regained the ability to pull the bar of the testing apparatus with measurable force in comparison to the MVC group at WPO8. In regard to the staircase test, the same authors reported a signi cant difference (p < 0.05) in favor of the ANG group at WPO4 while there were no signi cant differences detectable at WPO8 and WPO12 18 .

Electrophysiological evaluations
Electrophysiological evaluations including nerve conduction velocity and the amplitude of the evoked compound muscle action potential (CMAP) were performed in 3 studies and a total of 87 rats.

Median nerve model
One study 18 assessed the CMAP amplitudes of the thenar muscles by means of minimally invasive measurements in 16 rats every four weeks following median nerve resection and repair. At WPO4 and WPO8, respectively, no signi cant differences were detectable between the ANG and MVC group whereas CMAP amplitudes were highly signi cantly (p < 0.01) larger in rats treated with ANGs at WPO12. In addition, CMAP amplitudes in the ANG groups were highly signi cantly (p < 0.01) increased at WPO12 compared to WPO8. However, this difference was not signi cant in the MVC group.

Sciatic nerve model
Two studies 17,20 evaluated electrophysiological outcome metrics after sciatic nerve injury and repair in a total of 71 rats. Stossel et al. found that CMAP amplitudes recorded from the tibialis anterior muscles of MVC-treated animals were highly signi cantly (p < 0.001) smaller than those of the ANG group at DPO60, DPO90 and DPO120. While CMAP amplitudes in the ANG group were highly signi cantly larger at DPO90 (p < 0.001) and DPO120 (p < 0.01) in comparison to DPO60. This difference was not signi cant in the MVC group 17 . The authors also calculated the estimated axon loss based on their electrophysiologic recordings and found signi cant (p < 0.05) lower numbers of axons in the MVC group compared to the ANG group at DPO60. This difference became highly signi cant (p < 0.01) at DPO90 and DPO120, respectively. Additionally, Stossel et al. also performed the same measurements on the plantaris longus muscle. While they found no signi cant group differences at DPO60, animals of the ANG group had signi cantly (p < 0.05) larger CMAP amplitudes at DPO90. This difference was highly signi cant (p < 0.001) at DPO120. While there was no signi cant recovery of CMAP amplitudes in the MVC group at DPO120 compared to earlier time points, this recovery was highly signi cant compared to DPO60 (p < 0.001) and DPO90 (p < 0.01) in the ANG group. In regard to the estimated axonal loss, this was highly signi cantly more pronounced (p < 0.001) in the MVC group compared to the ANG group at DPO90 and DPO120. While a highly signi cant (p < 0.001) increment in axon numbers compared to baseline was observable in the ANG group at DPO90 and DPO120, this did not occur in the MVC group. Ramli and colleagues did not detect signi cant differences regarding CMAP amplitude, CMAP onset latency and nerve conduction velocity between the MVC and ANG group 20 .

Histological evaluations
Seven of the eight included preclinical studies (median nerve = 2, sciatic nerve = 5) with a total of 173 rats featured histological analysis of the reconstructed nerve, including 1) number of axons, 2) diameter of axons, 3) mean density as well as 4) size of myelinated bers and 5) thickness of the myelin sheath.

Median nerve model
Papalia et al. found no signi cant differences regarding the total number, ber diameter or myelin thickness between the ANG and MVC group at 6 months postoperatively 24 , whereas Stossel's group reported signi cantly (p < 0.05) and highly signi cantly (p < 0.001) lower numbers of myelinated axons in the MVC group at WPO8 and WPO12, respectively. The same authors observed no signi cant differences regarding the repaired nerves' mean cross sectional area, axon diameter, ber diameter, g-ratio or myelin thickness between groups, but found a highly signi cantly (p < 0.001) increased nerve ber density in the ANG group compared to the MVC group at WPO12 18 .

Sciatic nerve model
Brunelli and colleagues, who bridged 10 mm and 20 mm nerve defects of the sciatic nerve in their study, found a signi cantly increased number of axons in the distal nerve stumps of animals treated with MVCs when compared to the ANG group in both settings 11 . Geuna et al. assessed the total number, mean size and ber density of myelinated nerve bers and found no signi cant differences between the two groups 25 . Another study which evaluated histological parameters at DPO120 found that the total number of bers was signi cantly (p < 0.05) lower in the MVC group compared to the ANG group whereas there were no signi cant differences regarding nerve ber density, axon diameter, ber diameter, g-ratio and myelin thickness 17 . Ramli and colleagues assessed the 1) number of nerve bers, 2) degree of angiogenesis, 3) in ltration of immune cells, 4) in ltration of muscle cells and 5) development of traumatic neuroma. These authors did not report results of any statistical comparison between groups but found a higher number of nerve bers within the distal stumps of nerves reconstructed with an ANG. While the degree of angiogenesis, immune cell in ltration and muscle cell in ltration was equivalent between both groups, abundant neuroma formation was apparent in nerves repaired with MVCs. No neuroma formation was observable in the ANG group 20 . Ulkur et al. evaluated the number of myelinated axons and the mean axonal diameter at WPO28, reporting signi cantly (p < 0.05) higher numbers in regard to both counts in the ANG group 22 .

Muscle weight
Stossel's group assessed muscle weight of both the tibialis anterior muscle and gastrocnemius muscle at DPO120 following sciatic nerve repair, which was signi cantly increased in the ANG group compared to the MVC group 17 .

Clinical studies
Manoli's group found no statistically signi cant differences in regard to static or moving two-point discrimination as well as the Semmes-Weinstein-Mono lament test after digital nerve repair (n = 31) with either an ANG (n = 14) or a MVC (n = 17) 12 . Ahmad et al. also found no statistically signi cant differences in regard to two-point discrimination or muscle power assessed by means of the Medical Research Council scale following nerve repair (n = 18) with either ANGs (n = 9) or MVCs (n = 9) 16 .

Meta-analysis
Our meta-analysis revealed marked differences regarding the outcome of nerve repair by means of ANGs or MVCs between preclinical and clinical studies.   (Fig. 5). However, there was a slight trend towards better sensory function in the MVC group.

Discussion
Reconstruction of nerve defects in which tension-free primary neurorrhaphy is impossible requires interposition of an adequate guiding structure to facilitate nerve regeneration. However, use of ANGs, despite their status as gold-standard, requires careful consideration, regarding their limited availability and donor site morbidity following harvesting. The ideal nerve conduit should be abundantly available throughout the body, with high biocompatibility and -degradability in addition to optimum biomechanical properties. The regenerating nerve within the conduit should be optimally protected from surrounding scar tissue and in ltration of in ammatory cells, while its regeneration is maximally supported by the environment within the conduit 7,26 . MVCs were proposed almost three decades ago by Brunelli, who suggested that they characteristics match most of the aforementioned requirements for an ideal nerve conduit 11 . Our review and meta-analysis revealed marked differences between clinical and preclinical data in regard to the reported e cacy of MVCs and ANGs for reconstruction of peripheral nerve defects. Additionally, due to strong heterogeneity in study methodology, only a limited number of studies' data (n = 6) could be included in our quantitative synthesis. This was despite the fact that 10 studies were eligible for inclusion in our qualitative synthesis. Another main nding of this work is the overall limited number of clinical studies evaluating patient outcome following nerve repair with either ANGs or MVCs.
Preclinical studies reported diverging results with especially recent work pointing towards inferiority of MVCs compared to ANGs. Clinical studies, although limited in numbers, did not support these ndings, and showed non-inferiority of MVCs for reconstruction of upper extremity nerve defects. In order to elucidate the underlying reasons for these observations, some important methodological aspects of the included studies need to be addressed. Early preclinical studies which a rmed the high potential of MVCs, featured either a very limited number of animals, e.g. n = 3 11,25 or reported only partial results of the respective histological and functional assessment 11,21 . Included studies which were published more recently used a broader array of evaluation methods and group size was at least n = 8 17,18 . Additionally, superiority of MVCs in earlier studies was almost exclusively related to axon numbers in the distal nerve stump, which were signi cantly higher in the MVC group. No functional outcome was reported related to the histological assessment on the one hand while the majority of studies which assessed functional outcome reported inferior results for the MVC groups. It is of particular interest in this context that other authors have already emphasized an apparent discrepancy between axon numbers and the degree of functional recovery following experimental peripheral nerve injury and recovery 27 . Since nerve regeneration consists of the three main phases of 1) axonal regrowth 2) end-organ reinnervation and 3) functional recovery 28 care must be taken not to judge the overall outcome based on results of only one of these three distinct phases. As pointed out previously, comprehensive analyses of functional recovery were primarily published in recent works. Therefore, there is a growing body of literature indicating inferiority of MVCs to ANGs in the preclinical setting of murine models. Out of the eight studies included, three featured a "critical size" nerve defect, which is considered to be at least 15mm in rat models 29 . Two of these studies were published within the last 3 years and featured at least two assessments for functional recovery each. Notably, all three studies reported signi cantly inferior results for the respective MVC group, emphasizing apparent inferiority of MVCs to bridge long nerve defects. Clinical evaluations of MVCs were performed in a number of studies without an ANG control group 10,30−32 . Only two studies compared the functional outcome in comparison to ANGs. These studies reported no inferiority of muscle-in-vein conduits, which brings up the question what could be the underlying reasons for these observed discrepancies between preclinical and clinical data. In our opinion three major points must be addressed in this context. While the rst and second focus on surgical aspects of nerve repair, the third is based on neurobiological considerations. In regard to the rst point, we would like to emphasize that the way how the MVCs are prepared and coaptated to the nerve stumps is of crucial relevance for the outcome following nerve repair. Tension-free nerve repair has been emphasized by Millesi as absolutely paramount prerequisite for optimal nerve regeneration 33 due to the devastating consequences of tension on intraneural perfusion 34 . However, given the different preconditions of preclinical VS clinical studies of peripheral nerve repair, there might also be differences in regard to the way the respective nerve graft is coaptated to the nerve stumps. In a clinical situation, the patient's nerve has already been severed, which means in case of a segmental damage it is highly likely that this gap cannot be bridged by means of the original nerve. Therefore, if autologous tissue, e.g. a MCV or ANG, is placed between the nerve stumps, the surgeon can prepare the respective transplant according to the gap length to guarantee tension-free coaptation. On the other hand, in preclinical studies the nerve gap is created by the experimenter and needs to be of standardized length to guarantee comparability of experimental results. In consequence, the nerve gap is usually bridged with the original nerve in retrograde fashion as an autograft. However, given the tendency of the nerve stumps to retract the original nerve gap increases by < 3 mm, e.g. from 10 to 12 mm 35 . Therefore, placing the original autograft tension-free between the nerve stumps might be more di cult. In addition, we assume that most experimenters prepare a MVC of the same length as the nerve autograft, possibly hampering tension-free coaptation of this graft as well. In line with these considerations, this might complicate another crucial aspect of nerve repair by means of MVCs, which is pulling the nerve stump inside the vein rather than just coaptating the nerve's epineurium to the wall of the vein 20 . If the epineurium is sutured to the vein wall without pulling the nerve inside the conduit, regenerating axons might get trapped outside hindering the process of nerve regeneration 12 . Second, clinical studies on peripheral nerve reconstruction most commonly feature injuries of digital nerves, which are relatively small and monofascicular 36 . In opposite, mixed (motor and sensory) multifascicular nerves are most commonly studied in preclinical animal studies. Misdirection of axons, which has been shown to signi cantly impair recovery of target organ function 37 is therefore more likely to occur when a gap of a mixed nerve is bridged with an MVC, which does not feature the fascicular pattern of the original nerve. This is also supported by the results of a study reporting non-inferiority of MVCs compared to primary neurorrhaphy in a rat model of distal facial nerve injury, which is mostly comprised of muscular nerve bers at this level, hence the consequences of axonal misdirection might be less devastating for functional recovery 38 . The third important aspect which needs to be considered, relates to the process of neuroregeneration itself, which, despite striking similarities, is markedly different in certain aspects between rodents and humans 29,39 . This is especially relevant when nerve conduits are used as in the case of MVCs. The spatiotemporal course of nerve regeneration has been elucidated by several authors during the last decades and can be further subdivided in three distinct phases which are 1) the molecular and cellular phase, 2) the axonal phase and 3) the maturation phase in which Schwann cells are reprogrammed to adapt a myelinating or nonmyelinating phenotype 40 . In order for nerve regeneration to occur cellular debris and other potential obstacles for regenerating axons must be removed during the molecular and cellular phase following the process of Wallerian degeneration 41 . Then a brin matrix is formed between the proximal and distal nerve stump to act as a guiding scaffold for regenerating axons de ning the maximum length of axonal regeneration without interposition of a guidance structure, e.g. the "critical size defect". This critical size defect's length is approximately 1.5 cm in rats whereas it measures ~ 4 cm in humans 29 . One can visualizes the process of peripheral nerve regeneration through a conduit, e.g. an ANG or an MVC, as the simultaneous proceeding of degeneration/resorption of hindering tissue debris and regrowth of regenerating axons through the aforementioned brin cable. It has been established that longitudinally oriented bers, e.g. collagen or muscle bers, can support axonal regrowth and function recovery 11,43,44 . However, on the one hand, nerve regeneration is hindered if these guidance structures are degraded completely before the regenerating axons reach them. On the other hand, the bers must be completely degraded before nerve regeneration is completed, otherwise this will also impede complete recovery 45 . In case of an MVC the conduit is lled with muscle bers, consisting about 1/5 of protein 46 . These are degraded while simultaneous axon regrowth through the brin cable occurs.
Since the degrading processes of the muscle bers is mediated by proteases, it is of particular interest to consider the inter-species differences in regard to these enzymes. As shown for mice, rodents possess about 1.4 x the number of proteases and have a markedly increased proteome turnover in comparison to humans 47 . This is thought to be related to the shorter lifespan of rodents with their body function optimized to regain optimal integrity and function in the fastest way possible rather than maintaining the body's function over a long time 48 . We therefore hypothesize, that the degradation of muscle bers inside the vein conduit occurs markedly faster in rodents than in humans. As it was shown that an MVC collapses as soon as all muscle bers within have been degraded and this markedly hinders axonal regeneration 17 we hypothesize this to be one of the reasons for the observed inferior results of MVC nerve repair in rodents in comparison to ANGs. As human proteome turnover is slower compared to rodents, the muscle bers are also degraded slower, allowing more time to pass for the regenerating axon from the proximal stump to research the distal part of the MVC before it collapses. In conclusion, we suggest that this is a perfect example for profound differences between the human and rodent species which aggravates comparability of results in studies of peripheral nerve regeneration, especially in regard to critical size defects. However, critical size nerve defects represent a common clinical problem with a high need for adequate treatment options 29,39 . Use of MVCs to reconstruct peripheral nerve defects exceeding 3 cm has only been reported anecdotally and no concise conclusion is possible in regard to their feasibility for this scenario 49 . Although promising results for MVCs were obtained in clinical studies published until now, there is evident need for larger, clinical trials to gather more data regarding their use. Based on the results of this work we see a strong need to investigate their use in the setting of a critical size nerve defect in a non-rodent model. The disadvantages of rodent models, although cost-e cient and commonly used, have been pointed out by us. Large mammalians models such as sheep or pigs come with considerably higher costs as a disadvantage 50 . However, the results of these animal models, in contrast to rodent models, are more easily transferable to human patients 51 . Therefore, we suggest that MVCs should be evaluated in such large animal models before further clinical studies involving mixed motor nerves or nerve gaps of critical size should be intended.

Conclusion
Our results underpin that preclinical studies in rodents are not able to adequately prognose the feasibility of MVCs for reconstruction of critical-size defects in humans. This can be explained by differences regarding the experimental settings, and most important, profound inter-species differences in neurobiology. Additionally, data from clinical trials, in particular involving defects of mixed nerves, are still very sparse. Although large prospective human studies to answer these questions will be inevitable, we strongly advise for large animal studies beforehand. Our work emphasizes two main problems of peripheral nerve research: limited translatability of results from rodents to humans and profound differences regarding study methodology. While MVCs tackle several main problems of peripheral nerve repair by means of an abundantly available, autologous material with high biocompatibility, their promising features have not yet been explored in enough detail to support their unfettered use for nerve repair in human patients.

Methods
According to the PRISMA statement and in adherence with the recommendations found in the literature 52 56 . Sensory recovery can be evaluated by means of the "Pinch Test", in which toothed forceps are used to nip the toes of the respective paw. Besides the rat sciatic nerve model, the median nerve model is another frequently used model in preclinical research 57 .
Voluntary motor function in this model is most commonly assessed by means of the grasping test. While being held at its tail, the rat grasps a bar connected to a scale to measure and record the maximum force exercisable by the nger exor muscles. As these are predominantly innervated by the median nerve in rats, this method allows to determine the degree of functional recovery after median nerve reconstruction.
The staircase test is an alternative method to assess motor function following median nerve injury and reconstruction. It evaluates the ne motor skills of the forelimb while the examined rat retrieves spheroidshaped pellets of food from a staircase apparatus with restricted latitude 18 . As secondary outcome, electrophysiological parameters such as the nerve conduction velocity (NCV) or the compound muscle action potential (CMAP) of the respective target muscles were evaluated. Muscle weights of target muscles were chosen as complementary outcome metrics. Additionally, histological and histomorphometric parameters, e. g. the diameter of myelinated axons, the myelin sheath thickness, gratio and the density or number of myelinated nerve bers distal to the lesion site or inside the MVC and ANG were assessed. Regarding the clinical studies included in our synthesis, recovery of sensory function as assessed by static two-point discrimination was chosen as primary outcome parameter. Additional outcome metrics were moving two-point discrimination, the Semmes-Weinstein-mono lament test which is used to assess cutaneus sensory thresholds by applying nylon laments of different bendability to the respective skin area, and the Medical Research Council scale 58 for muscle strength. The methodological quality of all included studies was assessed and the risk of bias was judged in accordance with the recommendations of the Cochrane Handbook for Systematic Reviews of Interventions 59 for the following aspects: random sequence generation and allocation concealment (selection bias), blinding of participants and personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias), selective reporting (reporting bias) and other bias, e.g.
suspected insu cient statistical power or methods or use of non-standardized or strongly modi ed experimental methods.  Preclinical studies were clustered into two groups, either featuring the sciatic nerve or median nerve model. Further statistical subgroup analysis was based on this sorting. However due to a strong heterogeneity regarding the assessed outcomes, length of the observation period and surgical methodology, a meta-analysis could only be performed on the sciatic nerve model group. We used the inverse variance method in a xed effect analysis-model and expressed the results as difference in means (DM) for continuous outcomes with 95% con dence interval (CI) 60 . Since the Review Manager requires the standard deviation for each outcome, but some studies reported the standard error of the mean (SEM) instead, the Standard Deviation was calculated by multiplying the square root of the sample sizes with the SEM. The weight of the estimated intervention effect of each individual study was then determined based on the width of the con dence intervals of the respective study. In conclusion, studies with narrower con dence intervals had a higher impact on the overall average effect of interventions for the respective reconstructive approach (MVC or ANG respectively) 61 . Heterogeneity among studies was assessed by means of the Chi 2 and I 2 test 62 .

Declarations
Competing interests