13

Nerve grafts
Michel Merle and Aymeric Lim

Repair of nerve defects by grafts was first attempted by Phillipeaux and Vulpian in 1870
and then by Albert (1876). However, later it was abandoned because of inconclusive
results. It was not until 1972 that Millesi validated the fascicular nerve grafting method of
repair by reporting good results (Millesi et al 1972). In 1974, the first vascularized nerve
graft performed by Taylor seemed to provide a solution for clinical situations with long
segmental losses and poorly vascularized beds unfavourable for the take of conventional
grafts (Taylor and Ham 1976). Over time, however, vascularized grafts have not proved
their superiority over conventional grafts and their early promise has not been borne out
due to the scarcity of suitable donor sites.
The reconstruction of long segment peripheral nerve defects or brachial plexus injuries
requires the use of long grafts. The difficulty in finding donor sites of sufficient length
motivated us to experiment with vascularized allografts in the rabbit (Bour and Merle
1989). Cyclosporine was used during the period of nerve regeneration. Stopping the
immunosuppressive treatment inevitably led to rejection of the graft. Mackinnon
preferred to use a method which involved pretreating the allografts to minimize rejection
after cessation of the immunosuppressive treatment (Mackinnon et al 1984). The first
allograft hand transplantations have shown that it is possible to recover sensory and
motor functions aided by a combination of immunosupressants. Among these, is
tacrolimus which, while promoting nerve regeneration, also unfortunately causes
neurotoxic side effects.
Lundborg first demonstrated that it is possible for a nerve to regenerate in a silicone
chamber (Lundborg and Hansson 1981). The concept of nerve tubes was further
developed by Restrepo et al (1985b), Madison et al (1988), Fields et al (1989),
Mackinnon and Dellon (1990a), and Archibald et al (1991). It is now accepted that an
artificial or biological tube is able to sustain nerve regeneration over a distance of less
than 15 mm.
Before the advances in microsurgical techniques led to the application of the principle
of direct primary suture, nerve grafting was considered to be the only option in the repair
of peripheral nerve lesions. Currently, no well conducted study has been able to prove
that nerve grafting, when possible, is superior to primary or secondary suture. Grafts are
thus indicated solely for the repair of nerve gaps.

Different types of graft

The choice of graft is important because it directly affects the results of the repair. The
repair of digital nerves requires similar sized nerve grafts harvested from the medial
 Severe traumatic defects of the upper limb 168

antebrachial cutaneous nerve and more infrequently, vascularized grafts. For defects in
larger nerve trunks, sural nerve grafts are the most appropriate. A non-vascularized trunk
graft may be chosen when the decision has been made to sacrifice a major nerve. There
are problems with the vascularization of this volume of nerve tissue, however, and when
axonal regrowth occurs within the graft, it is often necrosed and sclerosed in its centre. It
is preferable to transform the nerve trunk into fascicles by intraneural dissection. Trunk
grafts are only indicated if vascularized. Strange (1947) was the first to propose the
concept of a two-stage repair of the median nerve with a vascularized ulnar nerve and
Taylor and Ham (1976) were the first to perform a free vascularized nerve graft using
microvascular sutures.
The concept of fascicular nerve grafting is impractical as, for the median nerve, it
would entail the use of one fascicle each for each of the 1620 fascicles that compose the
nerve.
All our laboratory experience with nonvascularized and vascularized allografts has not
been applied clinically because of poor results. We have demonstrated that it is possible
to reestablish the continuity of a nerve over a short distance with an empty tube of
perineurium (Restrepo et al 1985b). For gaps of less than 15 mm, silicone,
polypropylene, collagen and polyglycolic acid tubes have been used. Despite numerous
appeals in favour of the role of tropism in this technique, we remain convinced that the
best method for nerve repair lies in the improvement of the technique of direct
coaptation.

Fascicular grafts

This is the most commonly used graft for the repair of nerve defects.

The choice of donor nerve

The sural nerve (Fig. 1)

We usually use this nerve. It is harvested from distal to proximal. A 6 cm retromalleolar
incision localizes the terminal two branches of division. A second transverse incision 16
cm proximal to the lateral malleolus (as described by Gilbert et al 1986) allows extraction
of the distal part of the sural nerve as well as identification of its junction with the
accessory sural nerve originating from the common peroneal nerve. Harvesting both
nerves is useful for large defects or the simultaneous repair of multiple nerves.
In young children, harvesting should be done via a single incision because of the
abundance of fatty tissue and fascia surrounding the nerve.

The medial antebrachial cutaneous nerve

For small defects of digital nerves, we prefer to harvest this nerve from the distal aspect
of the arm. An oblique incision is made from the hollow on the medial aspect of the
biceps tendon
 Nerve grafts 169

Figure 1
Technique to harvest the lateral saphenous nerve and its
accessory branches. (a) The sural nerve arises from the
tibial nerve at the level of the popliteal fossa where the
superficial peroneal nerve also arises which takes its origin
on the common peroneal nerve and which travels
subcutaneously. (b) It is always necessary to perform an
additional incision at the junction of the top and middle
third in order to harvest both branches of the sural nerve.

to 5 cm proximal to the medial epicondyle. The basilic vein is retracted and anterior and
posterior branches of this nerve are exposed.

The lateral antebrachial cutaneous nerve

We have stopped using this nerve, the terminal branch of the musculocutaneous nerve, as
there is significant morbidity from scarring and loss of sensation at the donor site. It is a
donor nerve of last resort.
 Severe traumatic defects of the upper limb 170

Technique
The technique of nerve grafting follows some general principles: the incision should be
some distance from the nerve; it should be extensile and allow excision of scar tissue; it
should allow transfer of healthy tissue to improve the vascularity of the bed; and finally,
it should permit the simultaneous repair of any major vessels.
There are two methods of grafting. The classically described method is that of Millesi
et al (1972), which involves the preparation of a fascicular group and the suture of every
fascicle within. The second technique, which we learned from Narakas, focuses on the
nerve trunk. A monobloc of fascicles with similar diameter to the nerve to be grafted is
assembled with tissue glue. The nerve should be resected until healthy nonindurated
tissue is obtained so as to minimize a fibroblastic reaction. Meyers instrumentation is
best for the secondary resection of a nerve.

The technique of Millesi (Fig. 2)

The nerves are resected until healthy tissue is obtained. Methylene blue applied to the
nerve

Figure 2
Fascicular graft according to Millesi. After a partial
epineurectomy and trimming of the fascicular groups, the
fascicular grafts are adjusted and sutured with 10/0 nylon.

ends then reveals the fascicular arrangement. Rarely monofascicular or oligofascicular

but most often polyfascicular, they are arranged in groups or independently. If the
organization is monofascicular or polyfascicular, multiple nerve segments are apposed to
the large diameter fascicle. In polyfascicular nerves, one graft is coapted to one fascicle.
Grafting of mono-or oligofascicular nerves does not necessitate intraneural dissection
or resection of the epineurium. For polyfascicular distributions, however, Millesi
recommends staggering the repair zones so as to avoid having all the coaptations in the
same plane. Although apparently logical, this method may be injurious to the nerve;
resecting the epineurium alters the blood supply and dissociating the fascicles increases
the risk of fibrosis and collagen invasion. Finally, staggered resection of fascicles can
only be done with curved scissors with consequent crushing of the nerve and a natural
evolution towards necrosis.
 Nerve grafts 171

Each nerve segment is fixed with one or two sutures of Ethilon 9 or 10-0 while trying
to respect the correspondence between the proximal and distal nerve segments, which is
not always obvious.
We were very impressed with the time Millesi spent while performing a nerve grafting
operation, drawing with India ink on a sterile cardboard, the fascicular organization of
nerve ends separated by a few centimetres. Intraneural dissection can create the illusion
of fascicular correspondence over a certain distance. But the work

Figure 3
Fascicular nerve graft of the median nerve at the level of
the carpal tunnel.

of Sunderland (Sunderland 1945, Sunderland et al 1959) describing the changes in

direction in intraneural fascicular topography over 45 cm segments destroyed the
illusion that nervous anatomy could be accurately restored by grafting. On the other hand,
Jabaley et al (1980) demonstrated that, over a significant distance in the same quadrant,
were to be found the same fascicular groups linked by numerous ramifications.
In summary, it may be concluded that one should always try to restore the general
orientation of a nerve so that grafting may reflect the correspondence of the quadrants. To
look for a correlation at fascicular level remains an illusion. One must note that there are
certain levels at which it is possible to find a very marked fascicular organization; for
example the radial nerve at the elbow joint where one perceives a distinct organization of
the fascicles before actual division into posterior interosseous nerve and superficial
sensory branch. It is also the case for the ulnar nerve at the wrist where one finds
separation of the fascicular groups destined for the interossei and for the skin.
In these situations, the anatomy allows efficient interfascicular grafting; any
topographical error would send the motor nerves into cutaneous territory and vice versa.
The course of the graft is not necessarily the shortest between the two nervous
extremities. It is sometimes necessary to elongate the graft so as to place it in a healthy
bed favourable for its revascularization. Thus, when one would like to avoid returning via
the palmar approach to a multi-operated digit, it is possible to re-establish the continuity
of the digital nerve by placing the graft in a dorsolateral course with sutures in healthy
tissue. The length of the graft should be calculated by placing the operated limb or digit
in maximal extension so as to avoid all tension at the suture sites.
 Severe traumatic defects of the upper limb 172

The technique of Narakas (Fig. 4)

This technique has the advantages of being simple, less invasive and fast whilst
respecting the major principles of reconstructive surgery.
Narakas resects the nerves out of the fibrotic zone with the aid of Victor Meyer
forceps. The nerve ends are stained with methylene blue and are primarily oriented using
the epineural vascularization as a guide and secondarily the fascicu

Figure 4
Fascicular graft according to Narakas. (a) The nerve defect
is measured by extending the neighbouring joints. (b) The
sural nerve graft cleaned from its connective tissue is fitted
so as to have the same diameter as that of the nerve to be
grafted; they are joined at each end with Tissucol. (c)
Trimming of the nerve graft is performed with a
guillotine clamp of V. Meyer. (d) The graft is jointed to
the nerve stumps with Tissucol.

lar distribution if the defect is only of a few centimetres. The gap is measured carefully
with the limb in maximal extension and recognizing that this distance is always longer
than the real nerve loss, 12 cm are added depending on the course of the graft.
The sural and, if necessary, the accessory sural nerve are harvested with multiple stab
incisions and then cleaned under the microscope of all fat and connective tissue leaving
only two or three fascicular groups. The sural nerve is then laid out on a block of
polyethylene and folded on itself until the aggregate diameter approximates that of the
nerve to be grafted. The ends are bunched together like a bundle of firewood, glued with
Tissucol (Baxter, USA) and trimmed with Victor Meyer forceps. The middle of the graft
is left free for the nerve segments to spread in the bed.
This technique avoids all intraneural dissection. Glue replaces suture material which
always causes a foreign body reaction. The graft also has the advantage of being made to
measure. It is, however, less precise with respect to fascicular groups and some nerve
segments may be wasted by coaptation with non-neuronal epineural tissue (Fig. 5). We
have used this technique satisfactorily since 1987. Our results are comparable with those
of Millesi with the added advantage of simplicity of technique. The use of glue does not
impose a barrier to nervous regeneration (Merle 1992).
 Nerve grafts 173

Figure 5
Ulnar nerve graft at the level of the wrist. (a) The nerve is
trimmed to a healthy area using the V. Meyer clamp. (b)
The sural nerve is fitted as the same diameter as the ulnar
nerve. (c) Both extremities of the nerve graft are glued
with Tissucol. (d) Trimming of the nerve graft using
freezing technique. (e) Gluing of the nerve graft to the
ulnar nerve using Tissucol.

Postoperative care
All grafts must be immobilized postoperatively. When the length of the grafts has been
calculated with the wrist in mild extension, it suffices to immobilize the wrist in that
position for 3 weeks. For digital nerves, the metacarpophalangeal joints are immobilized
 Severe traumatic defects of the upper limb 174

in 60 flexion for 3 weeks while allowing early mobilization of the proximal and distal
interphalangeal joints.

Indications
Grafts are indicated in all nerve defects which cannot be repaired directly without tension
despite the methods of mobilization, transposition and bone shortening. The natural
separation of a nerve after section is linked to its elasticity. This can range from 1 cm to 2
cm for a lesion at the wrist to 5 cm at the level of the arm. Mobilization of the nerve by
extensive dissection with mild flexion of the wrist allows primary repair in most cases. If,
however, the resection of a neuroma creates a gap of 4 cm or more in the wrist or the
forearm, grafting is indicated.
Narakas technique is the simplest and one must not hesitate to elongate the graft so to
as to allow a course in a healthier bed.

Results
Grafting of nerve defects is reserved for the most severe injuries. This explains the
mediocrity of our results (Dumontier et al 1990). Despite the introduction of
microsurgical techniques, in 10 years we have not observed any significant improvement
in useful results which have ranged from 23% to 25%. We find ourselves in conflict with
the multicentric study performed by Frykman and Gramyk (1991) who observed 81%
useful motor results and 79% useful sensory results after grafting. Superior to the best
results of primary repair, these figures are difficult to understand, especially given all the
known problems with grafting. It is, however possible that many of their patients may
have benefited from direct repair if they had been subject to our indications. Finally, the
superiority of their results may be due in part to the young age of the patientsless than
20 years old in 45% of the cases. We have demonstrated that after primary repair, useful
results were found in 88% of patients less than 10 years old (Merle et al 1984).
The accurate evaluation of results demands specificity for each nerve. The functional
result of the ulnar nerve needs to take into account the strength of grip, while useful
results for a median nerve would essentially reflect the return of sensory function. The
age of the patient, the nature of the injury and associated lesions are determinant factors
in the prognosis.

Vascularized nerve grafts

History (Table 1)
Large nerve defects, i.e. more 10 cm, have an uncertain outcome when they are repaired
by conventional fascicular free grafting techniques. Therefore, Taylor aroused great
interest when he described the use of a radial nerve to graft an extensive loss of substance
in a median nerve lesion (Taylor and Ham 1976). During the following decade, many
experimental studies were undertaken and this new technique was applied
 Nerve grafts 175

Table 1 Review of vascularized graft techniques.

Authors Year Technique
Reconstruction of the median by the pedicled ulnar in
Strange 1947
two procedures
Reconstruction of the tibial nerve by the pedicled
McCarty 1951
peroneal nerve
Taylor and Ham 1976 Superficial radial to median
Bonney et al 1984 Ulnar nerve in the forearm
Townsend and Sural nerve with its arterialized vein
1978
Taylor
Fachinelli et al 1981 Sural nerve and superficial sural artery
Comtet et al 1981 Medial cutaneous nerve
Ulnar nerve in the arm and proximal ulnar collateral
Lebreton et al 1983
artery
Oberlin et al 1985 Pedicled common peroneal nerve
Internal branch of the anterior tibial nerve and dorsalis
Rose 1985
pedis
Anterior interosseous nerve and anterior interosseous
Merle and Dautel 1991
artery

clinically. However, with the lack of suitable donor sites and results that did not appear to
be better than those with conventional grafts, enthusiasm and interest subsided.
The principle of the vascularized nerve graft goes back to Strange who described the
possibility of grafting the median nerve by staged transposition of the ulnar nerve; he
obtained his first success in 1948 (Strange 1947, 1948). The same principle of pedicled
grafting was applied by McCarty to repair the tibial nerve (McCarty 1951).
Microsurgery revived interest in the technique of nerve repair by vascularized
grafting. Taylor used the superficial branch of the radial nerve, 24 cm long, vascularized
by the radial artery to repair a defect of the median nerve. This technique is acceptable in
patients presenting multiple trauma of the limbs, because sequelae at the donor site are
minimal. However, the sensory loss would be unacceptable in the patient presenting with
only a single injury.
Taylor (1978) classified vascularization of peripheral nerves into five typesthe first
three of which could be used for free vascularized transfers (Fig. 6). Taylors work in the
field of donor sites was taken up by others. Bonney et al (1977) suggested use of the
antebrachial portion of the ulnar nerve. Comtet et al (1981) described the vascularization
of the internal cutaneous nerve and Fachinelli et al (1981) described the superficial sural
artery which vascularizes the sural nerve. Lebreton et al (1983) described the
vascularization of the ulnar nerve in the arm and Oberlin and Alnot (1985) described the
pedicled common peroneal nerve. Rose (1985) repaired digital nerve defects using the
internal branch of the deep peroneal nerve lifted with the dorsalis pedis artery: later, he
came to use only a vena comitans which he arterialized. This arterialization technique
was described by Townsend and Taylor (1984) for the sural nerve and was also
 Severe traumatic defects of the upper limb 176

developed by Gu et al (1985). Recently, Dautel used the anterior interosseous nerve taken
with the anterior interosseous artery (Merle and Dautel 1991).
Our clinical experience summarizes the problems encountered in the upper limb, i.e.
avulsions of the brachial plexus, extensive loss of substance with poor tissue environment
(Volkmann, crush) and defects of digital nerves.

Vascularized grafts for extensive loss of substance of the brachial plexus

In cases of intraspinal avulsion of C8 and T1 roots, it is legitimate to use the brachial or
antebrachial portion of the ulnar nerve to bridge an extensive gap, from C5/C6 to the
upper trunk or even to the posterior cord.
We used the brachial portion of the ulnar nerve in five cases, five times as a free
vascularized transfer and four times as a pedicled graft. By the use of free vascularized
grafts measuring about 16.5 cm, it was possible to bridge gaps averaging 14.4 cm in
length. Usually, vascular anastomoses

Figure 6
Five types of vascularization of peripheral nerve described
by I. Taylor. Types a, b, c, could be used for free
vascularized transfers.

of the proximal ulnar pedicle were made with one of the cervical or thoracic branches of
the subclavian artery.
Four out of the five vascularized transfers remained viable. The Tinel sign progressed
distally at an average of 3 mm per day in the first 6 months after operation. Contractions
of the biceps were first observed clinically and electromyographically in the ninth month
after operation whereas the same result would be observed 12 months after non-
vascularized fascicular grafts. The one observed failure was caused by early thrombosis
of the arterial and venous anastomoses and since extrinsic neovascularization of whole
nerve trunk grafts is poor or delayed, the grafted nerve underwent ischaemic necrosis.
Functional failure was complete in this case.
In the four cases where we used the pedicled brachial segment of the ulnar nerve, we
observed recovery of the biceps graded M3+ (three patients) and M4 (one patient). In this
 Nerve grafts 177

situation, there is no risk of thrombosis but the length of the graft is a major concern. It is
imperative to dissect the ulnar nerve as far as 6 cm distal to the medial epicondyle of the
humerus. This is necessary in order to mobilize the ulnar nerve upwards and suture it to
the C5 or C6 root without tension.
In this short series, it is worth noting the consistency of results of biceps function
when the vascular anastomoses remain patent. Yet, when the patients were examined
after 3 years or more, functional results were no better than those obtained with
successful conventional fascicular grafts; progress of the Tinel sign, which occurred
through the vascularized graft at above average speed during the first 6 months (3 mm
per day), later stabilized at 1.5 mm per day.
Our results were similar to those obtained by Bonney et al (1984) (two failures out of
12 cases), and Alnot (1988, one failure out of 10), who used the ulnar nerve with its
artery. We avoided using the antebrachial portion of the ulnar nerve so as not to sacrifice
the dominant artery of the hand.

Vascularized nerve grafts for extensive loss of nerve substance in the

forearm
Severe trauma of the forearm (crush and avulsion injuries, Volkmanns contracture)
necessitate the restoration of at least sensory function of the median nerve. The tissue
environment is usually poor and cannot support free fascicular grafts. Grafting the
median nerve with the vascularized brachial segment of the ulnar nerve seems to be a
good choice (Fig. 7).
We have used this technique on three occasions, two of these for Volkmannns
ischaemic contracture. The first two cases failed: early thrombosis of the vascular
anastomoses resulted in necrosis of the grafts and no functional recovery was observed.
This emphasizes the uncertainty of microsurgical techniques applied to

Figure 7
Ulnar vascularized nerve graft to repair median nerve at
the level of the wrist. (a) Sequelae from a median and an
ulnar nerve injury in a 60-year-old patient. Total
anaesthesia of the hand. (b) Design of the free compound
flap: the skin flap is vascularized by its ulnar artery which
also vascularized the ulnar nerve. (c) The free compound
flap. (d) Result after 1 year, protective sensitivity has
returned in the territory of the median nerve.
 Severe traumatic defects of the upper limb 178

recipient vessels severely damaged by the initial trauma.

Grafts of digital nerve

The published results of conventional digital nerve grafts vary widely. One can only
admire the results of Mackinnon and Delton (1988) who reported 31 cases of
discriminating sensitivity out of 33 operations, i.e. a 93% success rate. Tenny and Lewis
(1984) obtained discriminating sensitivity in 32 cases out of 42. In our experience,
Dumontier et al (1990) observed discriminating sensitivity in three out of 16 patients.
These disappointing results led us to investigate vascularized grafts. At first, we used
Roses (1985) technique: the internal branch of the deep peroneal nerve is harvested with
a vena comitans that is used to bridge the digital artery. Later, we switched to the
technique of Dautel using the anterior interosseous nerve and its artery (Merle and Dautel
1991). Our prospective study included nine patients. The average length of the graft was
35 mm. In the first five cases, the donor site used was the one described by Rose, i.e. the
internal branch of the deep peroneal nerve and the vena comitans adjacent to the dorsalis
pedis artery. In the last four cases, the segment of the anterior interosseous nerve distal to
the origin of the flexor pollicis longus nerve was used as a donor site. The vascular
anastomoses were end-to-end except in two cases where the proximal anastomosis was an
end-toside suture. Immediate patency was judged at tourniquet release with a patency test
done distal to both proximal and distal vascular anastomoses. All our grafts were found to
be patent. No further monitoring was done in the first days after operation.
Long-term vascular results were assessed using Allen test; in the first six cases, the
results of the test were compared with arteriographic data obtained in the third month
after operation. Since the results were consistently concurrent, no arteriography was done
in the last three cases. Thrombosis was observed in six of the grafts at the time of control;
out of the three remaining patent grafts, two were those cases where the proximal
vascular anastomosis was end-to-side. Assessment of sensory results was done with the
static and dynamic two-point discrimination tests. No selective block of the healthy
contralateral nerve was used to study the sensory result. However, care was taken to test
precisely the area of the pulp that was innervated solely by the grafted nerve. Sensory
results observed in these nine clinical cases were as follows: three cases were considered
failures since static discrimination was equal to, or worse than, 15 mm. No patient had a
static two-point discrimination better than 10 mm. Even though this series was too small
for statistical analysis, it seems that long-term sensation was not influenced by graft
patency.
The vascular results obtained in this series demonstrated the difficulties of late
reconstruction of digital vascular axes when the contralateral axis is patent. Since there
was no long-term postoperative monitoring, the exact time at which thrombosis occurred
is not known. The sensory results of the six cases with long-term thrombosis were not
different from those of the three cases where vessels remained patent. Thus, there is an
important difference between vascularized nerve graft involving small nerves, such as
digital nerves, and those involving large nerve trunks. When secondary thrombosis
occurs, small calibre nerves, such as the internal branch of the deep peroneal nerve or the
anterior interosseous nerve, are probably revascularized from their tissue environment as
 Nerve grafts 179

conventional non-vascularized nerve grafts would be. Our results are not in agreement
with those of Rose et al (1989).

Discussion
Our clinical experience of vascularized grafts indicates that the results, i.e. sensory and
motor functional recovery, were not significantly enhanced. This was true for both the
brachial portion of the ulnar nerve and small nerve trunks. The only guaranteed benefit,
observed in cases of brachial plexus lesion, was the consistent recovery of biceps
function at M3+ or M4 when the graft remained patent. However, when there was
thrombosis of the anastomosis, necrosis of the graft ensued and functional failure was
complete. This risk is not negligible in surgery of the brachial plexus and must be
weighed against the relative safety of conventional fascicular grafts.
The experimental work of Daly and Wood (1985) and of Lux et al (1988) on the dog,
demonstrated that blood supply of conventional nonvascularized grafts was superior to
that of vascularized grafts on day 4 to day 6, provided the tissue bed was healthy. This is
the case with brachial plexus lesions, where the tissue bed is usually satisfactory and
conducive to revascularization of free nerve grafts.
Claims of superiority of vascularized nerve grafts have been based mostly on
optimistic forecasts during the first month after operation, when the Tinel sign progresses
through the graft at a rate of 3 mm per day. This may be due to rapid phagocytosis of
myelin sheaths and enhanced activity of Schwann cells. The fact remains that
vascularized grafts have less tendency to sclerose than conventional grafts. The latter are
subjected to ischaemia for several days; although Penkert et al (1988) have shown that
rabbit Schwann cells can survive ischaemia for 67 days, there is little doubt that the
ischaemia hinders Schwann cell activity.
Axonal regrowth seems to be optimal in the vascularized graft. However, at the
proximal and distal suture sites and when the nerve divides into collateral branches,
axonal sprouts meet with the same obstacles as in conventional grafts. This may explain
why the final results are similar with both techniques.
These clinical results are not as good as the results of animal experimentation.
Restrepo et al (1985a), and Shibata et al (1988) demonstrated in the rabbit that the
number and diameter of fibres, along with the thickness of myelin sheaths, were greater
in vascularized than in non-vascularized grafts. Pho et al (1985) found no histological
difference between vascularized and non-vascularized grafts in the rat. All donor sites are
not equally good. The trunk of the ulnar nerve is better because it has both arterial supply
and venous return. This is not the case with grafts of the internal branch of the deep
peroneal nerve or of the anterior interosseous nerve, where a vena comitans is arterialized
but no venous return is reconstructed. In these conditions, the nerve cannot be considered
to be physiologically vascularized. It is probable that the nerve soon suffers from venous
stasis, leading to oedema or even thrombosis. This insufficiency of venous return
probably explains the disappointing sensory results we have observed after grafting
digital nerves with Roses technique. The donor sites that we have listed in Table 1 are
few in number.
 Severe traumatic defects of the upper limb 180

Even when nerves are of sufficient diameter and vessels are of appropriate calibre,
these sites are rarely useful; the avulsions of C8 and T1 roots that justify their use are not
a frequent occurrence.
The technical obstacles might be overcome in the future: at present, they restrict the
surgeons choices. We prefer to ameliorate the quality of the tissue bed rather than to
continue transplanting rare vascularized grafts. Giving a new surface to sclerosed tissue
beds through the use of free or pedicled flaps will guarantee rapid revascularization of
conventional fascicular grafts. This option is all the more justifiable in cases of complex
trauma; the associated procedures on bone and tendons also benefit from a satisfactory
tissue bed.

Nerve allografts

The successful experience of immunosuppressive treatment for heart, renal, liver and
now hand transplants along with extensive experimental evidence encouraged the use of
nerve allografts. Freezing grafts and adjuvant immunosuppressants allow application of
these techniques in humans but there are unresolved technical and ethical problems,
which have limited clinical usage to the present time.

History
The concept of nerve allografts is not new and it is interesting to note that the clinical
case done by Albert in 1885 preceded by 5 years the work of Forsmann (1898) who
performed a successful allograft in a rabbit. This was followed by numerous
experimental works until 1945, on dogs, cats and monkeys with results ranging from
failure to success. After a gap of 20 years subsequent work by Dos Gupta et al (1967),
Zalewski (1971), Pollard and Fitzpatrick (1973), Chung and Chung (1974), Comtet and
Revillard (1979) and Levinthal et al (1978) between 1967 and 1981 showed constant
failure in the rat and pig.
From 1982 onwards, Mackinnon et al (1984), Bain et al (1987), Evans et al (1994) and
others demonstrated that nerve allografts in the mouse only work if they are refrigerated
in Belzers solution (University of Wisconsin Cold Storage Solution, Evans et al 1999)
for a minimum of 7 days with small doses of cyclosporin A. They also proved that the
immunosuppressive effect of monoclonal antibodies prevented rejection of the graft and
allowed good regeneration of the nerve (Nakao et al 1995). Experimental work was also
done in the rabbit and the rat to determine the feasibility of vascularized allografts (Best
et al 1993). We evaluated brachial plexus repair in the rabbit and the dog with massive
vascularized allograft. These grafts, which were only preserved for a few hours in
Ringers solution before revascularization, did not undergo any refrigeration. The rabbits
were then treated with cyclosporin A and signs of regeneration were obvious during
immunosuppressive therapy, however when the treatment was stopped all the rabbits
developed a massive rejection reaction with loss of function (Bour and Merle 1989). Best
et al (1993) compared vascularized autograft and vascularized allograft with and without
immunosuppressive therapy in the rat. They proved that vascularized allograft gave
equivalent results when compared to vascularized autograft. On the other hand,
 Nerve grafts 181

vascularized allograft without immunosuppression was subject to an acute and massive

rejection, with vascular thrombosis possibly accelerated by direct contact with recipient
antibodies. Concurrently, we have developed the clinical use of vascularized autografts.
However, we realized after a few years follow-up that the functional results were not
superior to conventional non-vascularized autografts when the surrounding tissue bed
was healthy permitting revascularization of the grafts. The clinical results of vascularized
autografts have not been remarkable, which explains the small number of studies on this
topic (Merle and Dautel 1991).

Experimental data in favour of non-vascularized allografts

Numerous studies, mostly on the rat, rabbit and goat, have demonstrated the value of
graft pretreatment with and adjuvant immunosuppressive treatment with or without
monoclonal antibodies.

Pretreatment of grafts
Preservation at 5C in Belzers solution (University of Wisconsin Cold Storage Solution,
Evans et al 1999) for a period of 7 days does not diminish the number of Schwann cells
but reduces the immunogenicity. Lower doses of cyclosporin are thus required
(Mackinnon et al 1992). This period of preservation allows the recipient to be
progressively immunosuppressed, making allograft nerve grafting an elective operation
unlike other allograft procedures. In addition, the number of myelinated fibres during
regeneration and the conduction velocity are increased. This effect is particularly evident
when FK 506 (tacrolimus) is used as the immunosuppressor. Injection of recipient
Schwann cells into the allograft also contributes to the protection of the graft from
rejection.

Immunosuppressive treatments
The benefits of immunosuppression with cyclosporin A have been well demonstrated in
the rat (Bain et al 1987, Mackinnon et al 1992, Strasberg et al 1996). The allograft allows
axonal regeneration in the host to proceed. The number of donor Schwann cells
diminishes while the recipient Schwann cells colonize the allograft. Without
immunosuppression however the rejection reaction is very rapid and the allograft
becomes a fibrous cord obstructing any nerve regeneration.
Atchabahian et al (1998) have demonstrated that immunosuppression can be stopped
without detriment to neurological function as long as nerve regeneration has reached the
sensory and motor end organs. Generally, at this stage, recipient Schwan cells have
finished colonizing the allograft. Cases of rejection have to be detected early so as to
reinforce the immunosuppressive treatment. It is possible to salvage an allograft which is
being rejected if FK 506 is added within 15 days (Feng et al 2001).
 Severe traumatic defects of the upper limb 182

Monoclonal antibodies
The usefulness of monoclonal antibodies in antirejection therapy for organ
transplantation has been well demonstrated. In the rat, different combinations have been
shown to protect the allograft from rejection while permitting a decrease in the dose of
cyclosporin A. This is the case for ICAM-1, LFA-1 (Nakao et al 1995) and CD4
(RIB502) (Doolabh and Mackinnon 1999). They are thus useful when there are adverse
effects from the immunosuppressive therapy, allowing a decrease in dose while
protecting the graft.

FK 506 (tacrolimus) and nerve regeneration

Lyons et al (1994) showed that in cell cultures, FK 506 increases the rate of axonal
growth. Since then, numerous studies have confirmed this capability. It has also been
reported that weak doses of FK 506 on treated allografts allow better functional results in
allografts than in autografts (Doolabh et al 1999).
The accumulation of all this experimental data encouraged clinicians to apply the
technique of non-vascularized allografts in humans.

Mackinnons clinical series (Mackinnon et al 2001)

Mackinnon successfully performed the first case in June 1988; a sciatic nerve to posterior
tibial nerve defect in an 11-year-old boy was repaired using a 10-strand allograft 23 cm
long. After 2 years and 2 months of immunosuppressive treatment (cyclosporin A), the
patient recovered some sensation but no useful motor function. Between 1988 and 1998,
Mackinnon carried out seven allografts on four women and three men for the following
indications: three severe median nerve defects, two associated with ulnar nerve defects;
one radial nerve defect; one sciatic nerve defect and two posterior tibial nerve defects. To
facilitate revascularization of the allografts, the nerves were stripped into fascicular
groups and all fat was removed. They were also placed under the skin to monitor better
for rejection.
The mean age of the subjects was 15 years. The mean length of the allografts was 190
cm. In the first two patients only allograft nerve was used for repair, while the five
subsequent patients also benefited from autograft sural nerve.
The mean time of immunosuppression was 18 months. Five patients were treated with
cyclosporin A and two with FK 506. Immunosuppression was stopped 6 months after the
Tinel sign was detected distal to the allograft with objective signs of return of sensation
and muscular function. The immunosuppressive treatment was as follows:
cyclosporin A 200300 ng/ml or FK 506 515 ng/ml
azathioprine 11.5 mg/kg/day
prednisolone 0.250.5 mg/kg/day for 58 weeks.
One graft was rejected after 4 months. The rejection reaction did not however affect the
autograft which was partially grafting the ulnar nerve. Long-term follow-up verified a
return of protective sensibility in six patients, with a two-point discrimination of 3 mm in
 Nerve grafts 183

one patient who had grafting of the ulnar nerve. There was useful motor recovery in three
cases.

Discussion
Currently, the technique used by all microsurgical teams for repairing nerve defects is
interfascicular autografting. Useful functional results, however, rarely occur in more than
60% of the patients in any series (Millesi et al 1972). Though vascularized autografts
have been useful in large defects, their use is limited by the donor site morbidity and the
fact that the functional results are not significantly superior to conventional fascicular
autografts (Merle and Dautel 1991).
Indications for nerve allografts do exist as shown in Mackinnons series (2001); in 10
years, she found 10 suitable cases but only seven were operated upon. While the sensory
results are encouraging, the motor recovery has been poor but one should note that most
of the limbs involved in this series were destined for a total amputation. There are many
arguments in favour of allografts; they are certainly an elegant solution for grafting large
nerve defects. The problems associated with the immunosuppressive treatment are in part
resolved by the limited duration of 18 months on average. Rejection reactions, which do
not affect the concurrently placed autografts, are monitored better by placing the grafts
subcutaneously. Nervous regeneration definitely benefits from FK 506 (tacrolimus)
(Lyons et al 1994). The main disadvantage is the risk of infection with viruses or even
prions. Is it ethical to subject a patient to these risks in the hope of obtaining a purely
functional benefit? Long-term follow-up and a detailed analysis of the secondary effects
of the immunosuppressive drugs would clear any ambiguity about their use.
Even if Mackinnons experience is unique, she had the great merit of basing her work
on a large body of experimental work, systematically resolving all the problems
encountered between 1967 and 1981. She was able to overcome them by cold treatment
of the grafts for 7 days at 5C. It was also with better immunosuppression protocols and
in particular the addition of FK 506 that the first clinical cases were possible.
The caution observed by microsurgical teams with regard to allografts will diminish
when the principles of allograft nerve regeneration are better known, the adverse effects
of immunosuppressive therapy are diminished and the risks of viral or prion infection are
addressed.

Nerve tubes

Lundborgs work using silicone artificial regeneration chambers demonstrates the role of
trophism for nerve fibre regeneration (Lundborg and Hansson 1981). The silicone
chamber was also an excellent experimental tool for the evaluation of the maximal gap
that a nerve can bridge when both ends are fixed to such a tube. The concept of a
neurotube, a substitute for autografts, had already been proposed by Weiss in 1943. With
Restrepo, we evaluated in the rabbit the use of an empty perineural tube to successfully
repair sciatic nerve defects of 15 mm. These results had been confirmed in humans with
the repair of digital nerve defects (Restrepo et al 1985b). Since then, numerous
biomaterials have been used to make a neurotube. Dellon and Mackinnon proposed a
 Severe traumatic defects of the upper limb 184

polyglycolic acid tube then evolved to trimethylene glycolide carbonate (Maxon) and
finally experimented with collagen (Mackinnon and Dellon 1990b).
All these experimental studies demonstrated the capacity of the nerve to regenerate if
the gap is between 5 mm and 15 mm. Reported complications range from kinking to
complete rejection while observing excessive foreign body reactions when the absorbable
biomaterial used is polyglycolic acid. Merle et al (1989) have described a chronic neuritis
resulting from a fibrous sleeve formed within a silicone tube causing ischaemia to the
nerve. Lundborg has rejected this type of complication and justifies its use in humans by
reporting the success obtained in seven patients with median or ulnar nerve defects of 35
mm (Dahlin et al 2001). Archibald et al (1995) have proposed the utilization of collagen
tubes. The results obtained in eight monkeys in which a gap of 5 mm in 15 median nerves
and 1 ulnar nerve were repaired with such a tube have demonstrated a functional result
equivalent to conventional grafting or direct suture.
Other teams have looked for resorbable materials with low inflammatory potential.
Using a tube made out of poly-3-hydroxybutyrate (PHB), Young et al (2002) reported
nerve regeneration in a rabbit over a distance of 4 cm. The use of a vein filled with
muscle (Fornaro et al 2001) encourages the migration of Schwann cells, the indispensable
guide to nerve regeneration. The work published by the Italian team of Battiston confirms
other similar works. The clinical application has been disappointing as the vein tends to
fibrose and sometimes becomes totally obstructed. The bed that hosts this composite graft
(vein and muscle) has to very good to support its revascularization. The accumulation of
numerous experimental works confirms that nerve regeneration in a neurotube is
possible over a distance of 515 mm. It is enhanced by using an inert material and a tube
cavity that facilitates the migration of Schwann cells. The integration, within the tube of
Schwann cells, of nerve growth factor and of resorbable filaments would probably
improve this concept.

Conclusions

The solution for the repair of nerve defects has not been found despite the multiple
therapeutic options. The fascicular autograft remains the best technique in most cases.
Vascularized nerve grafts have been disappointing and they should be reserved for use
only in cases involving very large defects and when the recipient bed is of poor quality
(crush injuries, Volkmanns ischaemic contractures).
It has not been proved that allografts with an immunosuppressive treatment that is
stopped after reinnervation can provide a useful functional result for the patient. In
addition, the ethics of using immunosuppressive drugs for a purely functional gain is
questionable.
For gaps of less than 15 mm, the utilization of a neurotube is justifiable but only time
will tell if it is better to use a biotube made with the patients own tissues or a tube
manufactured with resorbable materials and sown with Schwann cells and filled with
growth factor.