OUP user menu

The time course of epidermal nerve fibre regeneration: studies in normal controls and in people with diabetes, with and without neuropathy

Michael Polydefkis , Peter Hauer , Soham Sheth , Michael Sirdofsky , John W. Griffin , Justin C. McArthur
DOI: http://dx.doi.org/10.1093/brain/awh175 1606-1615 First published online: 5 May 2004


We sought to develop and validate a standardized cutaneous nerve regeneration model and to define the rate of epidermal nerve fibre (ENF) regeneration first in healthy control subjects and then in neuropathic and neuropathy‐free subjects with diabetes. Next, we assessed the effect of different factors on the rate of nerve fibre regeneration and investigated whether such an approach might offer insight into novel trial designs and outcome measures. All subjects had a standardized topical capsaicin dressing applied to the distal lateral thigh. ENF densities derived from skin biopsies were determined at baseline, after capsaicin treatment and at reinnervation time points. For each subject, the best fit line from post‐denervation data was determined and the slope was used as the rate of regeneration. In healthy control subjects, regeneration was correlated with psychophysical sensory testing, electron microscopy studies and immunohistochemistry with alternative axonal membrane markers. Topical capsaicin application produced complete or nearly complete denervation of the epidermis in both control subjects and people with diabetes. The rate of regeneration was associated with the baseline ENF density (P < 0.001), but not age (P = 0.75), gender (P = 0.18), epidermal thickness (P = 0.4) or post‐capsaicin treatment density (P = 0.7). ENF regeneration, as determined by recovery of ENF density, occurred at a rate of 0.177 ± 0.075 fibres/mm/day in healthy control subjects and was significantly reduced in subjects with diabetes (0.074 ± 0.064, P < 0.001) after adjusting for changes in baseline ENF density. Among subjects with diabetes, the presence of neuropathy was associated with a further reduction in regenerative rate (0.10 ± 0.07 versus 0.04 ± 0.03, P = 0.03), though diabetes type (P = 0.7), duration of diabetes (P = 0.3) or baseline glycated haemoglobin (P = 0.6) were not significant. These results have several implications. First, topical capsaicin application can produce a uniform epidermal nerve fibre injury that is safe and well tolerated, and offers an efficient strategy to measure and study nerve regeneration in man. Secondly, using our techniques, reduced rates of nerve regeneration were found in people with diabetes without evidence of neuropathy and indicate that abnormalities in peripheral nerve function are present early in diabetes, before signs or symptoms develop. These results suggest that regenerative neuropathy trials could include non‐neuropathic subjects and that trial duration can be dramatically shortened.

  • Keywords: diabetes; epidermal nerve fibres; outcome measure; peripheral neuropathy; regeneration
  • Abbreviations: DPN = diabetic polyneuropathy; EM = electron microscopy; ENF = epidermal nerve fibre; MDNS = Michigan Diabetic Neuropathy Score; PGP = protein gene product


Most neuropathies develop and progress slowly. If assessment of neuropathy is done based on pathology, nerve biopsy can be performed only twice, usually using the sural nerve, and has significant limitations. In clinical trials, repeated measurements of physiological function require assessments over months to years to determine the direction and speed of a significant change. In the context of clinical trials of regenerative agents, this translates into the need for long trial durations to detect any treatment effect on nerve fibre regeneration (Pfeifer and Schumer, 1995; Pfeifer et al., 1997).

We describe a novel method, the quantitation of regeneration of epidermal nerve fibres (ENFs) following a standardized chemical denervation, to study nerve regeneration. This technique may also be useful as an outcome measure in peripheral nerve trials.

ENFs are unmyelinated somatic nerve fibres that project to the skin and are ideal targets to study nerve regeneration. These fibres are easily quantified, can be measured repeatedly, and have been shown to be sensitive to capsaicin, which causes their degeneration followed by subsequent regeneration (Simone et al., 1998; Nolano et al., 1999). Lastly, these fibres are clinically relevant as they are implicated in many types of painful neuropathy such as those associated with human immunodeficiency virus (HIV) infection, diabetes mellitus and impaired glucose tolerance (McCarthy et al., 1995; Kennedy et al., 1996; Kennedy, 1996; Smith et al., 2001; Sumner et al., 2003).

Capsaicin is the active ingredient in hot chili peppers, and ENFs have been shown to be capsaicin‐sensitive nociceptors (Simone et al., 1998; Nolano et al., 1999). A previous study among five control subjects established that 3 weeks of daily topical capsaicin application produced ENF loss, followed by subsequent return of protein gene product (PGP) immunoreactivity (Simone et al., 1998; Nolano et al., 1999). The purpose of this study was to determine whether capsaicin application could be developed as a standardized model for measuring regeneration initially in normal volunteers, with expansion of the technique to include people with diabetes, with or without peripheral neuropathy. We sought to assess the effect of different factors on the rate of nerve fibre regeneration and to determine whether such an approach might offer insight into novel trial designs and outcome measures.


Thirty‐one healthy volunteers were screened for symptoms, signs and risk factors of peripheral neuropathy by history, examination and laboratory studies (HIV serology, vitamin B12, glycated haemoglobin, thyroid‐stimulating hormone, complete blood count and metabolic panel). Subjects were excluded from participation if they had a history of familial neuropathy or an abnormal screening test result.

Twenty subjects with diabetes were recruited from Baltimore area diabetes support groups. Subjects with diabetes included individuals with and without symptoms of neuropathy. A detailed peripheral nerve examination including clinical evaluation, nerve conductions (sural, median and ulnar sensory conductions and median and deep peroneal motor conductions) and skin biopsies at the distal leg, distal thigh and proximal thigh was performed. Laboratory data pertaining to glycaemic control were obtained by review of medical records. A Michigan Diabetic Neuropathy Score (MDNS) (Feldman and Stevens, 1994; Feldman et al., 1994) was calculated. The MDNS is a validated peripheral nerve examination scoring system in which 12 points can be assigned for sensory deficits at the toe, 18 points for power deficits in the hand and foot, and 16 points for reflex abnormalities. An MDNS score >6 and <12 corresponds to mild neuropathy (Feldman et al., 1994) and closely correlates with the Mayo Clinic‐derived NDS. Those with MDNS ≤6 were classified as being neuropathy free while those with scores >6 were classified as neuropathic. All studies were approved by the Johns Hopkins Medicine Institutional Review Board. All subjects gave signed informed consent.

In all subjects, a standardized area was demarcated on the distal lateral thigh (Fig. 1). An occlusive bandage measuring 35 × 50 mm, and containing 1.8 g of 0.1% capsaicin cream (Chattem Inc., Chattanooga, TN), was applied to the area for two consecutive 24 h periods. Punch skin biopsies (3 mm) were performed under lidocaine local anaesthesia, as previously described (McCarthy et al., 1995). Biopsies were obtained at baseline, immediately after the 48 h capsaicin period and at regular intervals up to 100 days. The biopsy tissue was fixed for 12–18 h in 2% PLP (paraformaldehyde–lysine–periodate) and cryoprotected overnight (20% glycerol/20% 0.4 M Sorrenson buffer) at 4°C, then serially cut at 50 µM intervals with a freezing microtome. Four randomly selected 50 µM sections were immunohistochemically stained using a free floating protocol with rabbit anti‐human polyclonal PGP9.5 antibody (Biogenesis 1 : 1200). All biopsies were taken 0.5 cm inside the treatment margin in order to reduce the possibility that reinnervation could occur by a process of collateral sprouting.

Fig. 1 The location of capsaicin application and of biopsy sites. Sensory testing was performed in control subjects within the region where biopsies were not performed.

Biopsies were quantified by two trained, blinded technicians as previously described (McArthur et al., 1998). Intra‐rater reliability previously has been shown to be excellent at 0.86–0.94 (McArthur et al., 1998). A random 10% sample of biopsies were counted by both individuals, and a comparison of the data using Kendall’s rank correlation showed excellent inter‐rater agreement (correlation coefficient = 0.794).

To confirm the loss of ENFs after capsaicin treatment, additional studies were performed in control subjects including psychophysical sensory testing for heat pain and mechanical detection threshold testing, electron microscopy (EM), and immunohistochemical staining with a membrane‐bound axonal marker, Gα0 (Neomarkers, Freemont, CA).

At the time of biopsy, heat pain thresholds (C and Aδ fibres) were determined using a peltier device with an 8 × 8 mm contact area (Thermal Devices Inc., Model LTS3). Starting at a baseline temperature of 35°C, the temperature was ramped upwards at a constant rate of 0.85°C/s. Subjects pushed a button when they felt a sensation of heat pain. The results of five separate trials, 1 min apart, were averaged. A maximum temperature of 55°C was set to prevent injury.

Mechanical threshold determination (Aβ fibres) was performed using the up/down method of Dixon (1980) using six von Frey monofilaments (NorthCoast Medical, Morgan Hill, CA) calibrated to deliver a force of 0.078, 0.196, 0.392, 0.686, 1.568, 3.92, 9.8, 19.6, 58.8 and 147 mN, respectively. Testing began with the 9.8 mN filament. If the subject failed to sense the filament, the next strongest filament was used. Conversely, when a filament was detected, the next lower filament was tested. The up/down test sequence was performed for four trials following the initial deflection. This process was repeated five times at each time point, and the results from the five repetitions were averaged to obtain a final value. A 50% mechanical detection threshold was calculated (Dixon, 1980). Results are expressed as a change from baseline values.

To determine if epidermal thickness affected the speed of regeneration, epidermal thickness was measured at 15 regular intervals in each of the four baseline skin sections, for a total of 60 measurements per subject in 16 subjects. This included eight subjects each from the diabetes and control groups. Pairs were matched for gender and age to within 3 years. All measurements were performed using a validated Bioquant software‐based protocol. The epidermal depth was defined as the distance from the skin surface to the dermal–epidermal junction. An average of the 60 measurements was taken as the final value. All measurements were performed by a masked technician.

Statistical analysis

Statistical analysis was performed using STATA 7.0 (STATA Corp, College Station, TX). For each subject, ENF density from post‐denervation data points was plotted against time. Linear regression was used to calculate the slope of the best‐fit line for each subject, and the slope was used as the rate of regeneration. Multiple bivariate regression models were then performed to determine the effect of individual predictors on the regeneration rate. The effect of diabetes, baseline distal thigh ENF density, post‐capsaicin application ENF density, age and gender on regeneration rate was examined in all 50 subjects as well as separately for healthy controls and subjects with diabetes. The effect of epidermal thickness was examined in 16 subjects. Possible interaction between pairs of significant variables among all 50 subjects was tested by combining variables in the linear regression. Variables reaching statistical significance at the 0.20 level in bivariate analyses were included in the final multivariate regressions using stepwise procedures. Differences in the regeneration rates between diabetic subjects with and without neuropathy were assessed through a bivariate regression model and using Mann–Whitney U test. Regeneration was also measured in relation to the baseline ENF density by plotting the percentage baseline ENF density against time for each subject. This was performed in order to demonstrate that the reduced regeneration rate in people with diabetes was not due simply to lower baseline ENF density values.


Of the 31 healthy control subjects enrolled, all completed the study. One subject with diabetes was removed because her baseline distal thigh skin biopsy was completely denervated. All subjects tolerated the capsaicin application and skin biopsies well. There were no infectious or bleeding complications, although some patients complained of localized, transient pain at the application site. The demographics of the subjects are given in Table 1. The subjects with diabetes were older (P < 0.001) and had lower baseline distal thigh ENF densities (P = 0.02) than healthy control subjects. There was no significant difference in post‐capsaicin ENF density among the groups.

View this table:
Table 1

Demographics of the study subjects

CharaceristicsDiabetesControls P
(n = 19)(n = 31)
Mean age ± SD (years)54.4 ± 13.737.5 ± 11.6<0.001
Male : female9 : 1013 : 18
Median duraton of diabetes (years)150
Glycated haemoglobin (mean ± SD)6.8 ± 1.15.4 ± 0.47<0.0001
On insulin pump (n)6
Type I/II diabetes10/9
Neuropathy (n)
 MDNS >68
 MDNS ≤611
Baseline (day –2) ENF density (distal thigh)16.6 ± 10.124.0 ± 10.60.02
 With polyneuropathy12.3 ± 6.3
 Without polyneuropathy19.6 ± 11.50.16
Post‐capsaicin application
 ENF density (day 0)0.23 ± 0.30.40 ± 0.40.06

Capsaicin application produced denervation of the epidermis, with no or only rare ENFs remaining after 48 h. Nerve fibres in the subepidermal dermal plexus were also sensitive to capsaicin treatment, and this region was also denervated. Reinnervation of the epidermis occurred in a stereotypic fashion over several months. The morphologies of epidermal and dermal nerve fibres from baseline and reinnervation time point biopsies were similar, and axons in the epidermis and dermis from regeneration time point biopsies were morphologically similar to baseline biopsies. Regeneration was not frustrated at the dermal–epidermal junction

In order to confirm that capsaicin application creates loss of ENFs which regenerate, we correlated PGP 9.5 results in control subjects with detailed psychophysical sensory testing, immunohistochemical staining with an alternative axonal marker, and EM imaging. Capsaicin application produced a mean 3.87°C increment in heat pain threshold temperature. Subsequent ENF regeneration was associated with a return towards baseline heat pain thresholds. There was no pattern of habituation among the five trials. In contrast, there was no change in mechanical detection thresholds with either denervation or reinnervation (Fig. 2). Dual staining of epidermal sections for both Gα0 and PGP 9.5 produced similar staining patterns (Fig. 3A), and adjacent skin sections stained with each marker alone yielded equivalent results. EM of baseline skin biopsies revealed axons in the subepidermal dermis and epidermis. After denervation, no axons were visible, consistent with denervation. A biopsy taken after 97 days of regeneration showed denervated Schwann cell bands with a central regenerating axon (Fig. 3B).

Fig. 2 Comparison of heat pain (A) and mechanical threshold (B) testing as measured from a change in baseline (day 0) values in healthy control subjects. Bars represent the 95% confidence interval. The mean baseline values for heat pain threshold were 46.26°C; 50.13°C at day 2; 49.13°C at day 14; 48.24°C at day 28; 48.67°C at day 44; and 47.56°C at day 55. The mean baseline value for mechanical touch detection threshold was 2.99 mN. Subsequent thresholds were 3.28, 2.99, 3.99, 4.63 and 3.94 mN on days 2, 14, 28, 42 and 56, respectively.

Fig. 3 (A) Immunohistochemical staining reveals co‐localization of PGP 9.5 (green in left panel) and Gα0 (red in right panel) consistent with Gα0 being an ENF marker. (B) EM through the subepidermal dermis (40 000×). The filled arrow indicates the margin of a reinnervated Schwann cell band, with the open arrow indicating the Schwann cell nucleus. The boxed area is shown enlarged with a regenerating axon.

For each subject, the slope of the line generated by plotting time against ENF density for post‐denervation time points was used as the rate of regeneration (Figs 4 and 5). The median R2 value for the regression lines was 0.88 (25% percentile 0.73, 75% percentile 0.95), indicating that the data were linear. The mean rate of ENF reinnervation among the 31 healthy subjects was 0.177 ± 0.075 fibres/mm/day. In bivariate linear regression analysis in which the effect of each predictor on the rate of regeneration was assessed, neither age (P = 0.80), gender (P = 0.1), epidermal thickness (P = 0.9) nor the post‐capsaicin application ENF density (P = 0.7) were significant. In contrast, the baseline distal thigh ENF density was significantly correlated with the rate of reinnervation (P = 0.02). Regeneration was sustained throughout the follow‐up period, although most subjects did not return completely to their baseline ENF density even after follow‐up periods as long as 100 days. Four control subjects followed for longer periods, ranging from 165 to 350 days, had sustained regeneration (data not shown).

Fig. 4 The reduced regenerative capacity of subjects with diabetes compared with normal controls by multiple different analytic techniques. (A) The data points for individual subjects. (B) Topographical plots of baseline and subsequent reinnervation ENF densities for each subject. (C) The reduced regenerative rate among subjects with diabetes with regeneration expressed as a percentage of baseline ENF density. The lines represent the mean rate of regeneration for healthy controls (left panel) and subjects with diabetes (right panel), P < 0.05. Mini‐graphs with an asterisk in A are from subjects with peripheral neuropathy.

Fig. 5 Raw data for 31 control subjects (A), neuropathy‐free subjects with diabetes (B) and neuropathic subjects with diabetes (C). For each subject, a regression line from post‐capsaicin time points is generated and the slope of this line is used as the rate of regeneration. The mean line for each group is shown as a red dashed line, and the rate of regeneration following denervation is 0.177 ± 0.075 (A), 0.10 ± 0.07 (B) and 0.04 ± 0.03 (C). The three groups of subjects have different rates of ENF regeneration. A versus B, P = 0.03; B versus C, P = 0.003; A versus C, P = 0.0001. The baseline time point for all three graphs is day –2, but has been shifted to the left for the purpose of demonstrating that denervation occurs in all subjects after capsaicin application. (D) The post‐capsaicin application regression lines for all subjects (dashed lines) and the mean regression lines for control subjects (solid red), neuropathy‐free subjects with diabetes (solid green) and neuropathic subjects with diabetes (solid blue).

Among the subjects with diabetes, eight had neuropathy with an MDNS >6. Subjects with or without neuropathy did not differ in age (54.4 ± 16.7 versus 54.5 ± 9.0 years, P = 0.99), duration of diabetes (14.9 ± 14.4 versus 20.1 ± 16.4 years, P = 0.47), gender distribution (55 versus 50% female) or baseline glycated haemoglobin values (6.97 ± 1.33 versus 6.46 ± 0.61%, P = 0.37). Baseline neuropathy measurements for all subjects with diabetes and stratified by the presence or absence of neuropathy are shown in Table 2. There was no difference in sural conduction velocity (P = 0.11), peroneal amplitude (P = 0.12) or baseline distal thigh ENF density (P = 0.16); however, sural amplitude (P = 0.003) and distal leg ENF density (P = 0.02) differed significantly.

View this table:
Table 2

The baseline electrophysiological results in all subjects with diabetes, and stratified by the presence or absence of peripheral neuropathy

MeasureSubjects with diabetes
AllWith DPNWithout DPN
(n = 19)(n = 8)(n = 11)
 Sural amplitude, mean ± SD (µV)10.5 ± 7.17.2 ± 5.916.2 ± 6.70.003
 (median, range)(11.1, 0–27.9)(5.4, 0–19.3)(13.5, 9.1–28.0)
 (normal: > 9 µV for age <60 years
 5 µV for age >60 years)
 Sural conduction velocity, mean ± SD (m/s)40.8 ± 3.740.4 ± 3.841.4 ± 3.80.11
 (median, range)(42.2, 34.9–46.8)(40.4, 34.9–45.8)(42.3, 35.5–46.8)
 (normal: > 39 m/s)
 Deep peroneal amplitude (µV)3.6 ± 2.83.7 ± 3.63.6 ± 1.40.02
 (median, range)(3.2, 0.1–12.1)(2.8, 0.1–12.1)(3.0, 2.0–5.4)
 (normal: > 2 µV)
Skin biopsy (IENF/mm)
 Distal leg (mean ± SD)11.5 ± 8.77.1 ± 5.316.7 ± 8.60.12
 (normative range: 13.8 ± 6.7)(9, 0.4–25.2)(5.7, 0.4–15.3)(14.5, 7–25.2)
Michigan Diabetic Neuropathy Score8.5 ± 9.917.5 ± 9.32.0 ± 2.20.0003

The P‐values reflect comparison of diabetes mellitus subjects with and without peripheral neuropathy.

Subjects with diabetes had a mean baseline distal thigh ENF density of 16.6 ± 10.1 fibres/mm, with capsaicin application producing near complete denervation (0.23 ± 0.33 fibres/mm). The baseline distal thigh ENF densities were lower (P = 0.02) in subjects with diabetes compared with healthy control subjects (Table 1).

The mean rate of reinnervation among the subjects with diabetes was 0.074 ± 0.064 fibres/mm/day and was significantly reduced compared with control subjects, P < 0.001 (Table 3, Fig. 4). This difference remained significant after adjusting for baseline ENF density and post‐capsaicin denervation ENF density, P < 0.001 (Table 3). The reduced rate of regeneration among the subjects with diabetes was also observed if regeneration was measured as a percentage of baseline ENF density (Fig. 4C).

View this table:
Table 3

Predictors of rate of regenerationMultivariate models
All study subjects (n = 50)β coefficientP‐valueβ coefficientP‐value
(95% CI)(95% CI)
(–0.117 to ‐0.026)(–0.113 to –0.038)
Baseline ENF density0.004<0.0010.004<0.001
(distal thigh)(0.002 to 0.006)(0.002 to 0.005)
Post‐capsaicin ENF density–0.01130.65
(distal thigh)(–0.061 to 0.03)
Male gender–0.0230.18
(–0.059 to 0.012)
Age (per decade)–0.00020.75
–0.016 to 0.012)

Regression analysis was performed on data from 50 subjects, of which 31 were healthy controls without signs, symptoms or risk factors for peripheral neuropathy, and 19 subjects had clinically confirmed diabetes. In bivariate modelling of all subjects, the effect of each predictor on the rate of regeneration was assessed. Only the presence of diabetes and baseline distal thigh ENFD were significant to P < 0.20 (data not shown). In multivariate modelling with all variables included, again only diabetes status and baseline distal thigh ENF density were significant (P < 0.005).

The final model included only these two variables, which remained highly significant (P < 0.001) even after adjusting for each other. In the final model, the β coefficients are interpreted as follows: the presence of diabetes is associated with a reduction in the rate of regeneration of 0.075 ENF/mm/day adjusting for baseline ENF density. For every 1 fibre/mm increase in the baseline ENF density, there is a 0.004 fibre/mm/day increase in the rate of regeneration after adjusting for diabetes status.

Among the subjects with diabetes, the regeneration rates varied considerably and prompted us to investigate the role of different predictors on regenerative capacity. Those with neuropathy had a reduced regenerative rate (0.04 ± 0.03 fibres/mm/day) compared with those without neuropathy (0.10 ± 0.07 fibres/mm/day), P = 0.03 (Fig. 5). There was no difference in the rate of regeneration in people with type I (0.069 ± 0.07 fibres/mm/day) versus type II diabetes (0.08 ± 0.053 fibres/mm/day). As with control subjects, age (P = 0.323), gender (P = 0.25) and epidermal thickness (P = 0.47) were not associated with an effect on regenerative capacity in bivariate regression analysis. The duration of diabetes (P = 0.304) and glycated haemoglobin level at the time of entry into the study (P = 0.618) were also not associated with the rate of regeneration. In contrast, baseline ENF density (P < 0.001) and the presence of neuropathy (P = 0.03) were associated with the rate of regeneration (Fig. 5).


In this study, we describe and validate a standardized model of chemical axotomy that reliably produces superficial denervation of the epidermis and subepidermal dermis in control subjects and people with diabetes. The results of this study have several implications. First, superficial denervation of the epidermis and subepidermal dermis is an efficient, safe and well tolerated strategy to measure and study nerve regeneration. Secondly, abnormalities in peripheral nerve function are present early in diabetes, well before signs or symptoms develop. Using our techniques, reduced rates of nerve regeneration were found in people with diabetes without evidence of neuropathy. While abnormalities in nerve regeneration among people with diabetes have been inferred from pathological studies (Bradley et al., 1995), our technique dynamically demonstrates reduced human nerve regeneration in a standardized nerve injury model. Thirdly, these results have implications for future regenerative diabetic polyneuropathy (DPN) trials.

Our goal was to develop a simple, reproducible model that was well tolerated and well suited to the study of nerve regeneration in humans. Previous studies have described epidermal denervation following either injection or repeated application of capsaicin over several weeks (Simone et al., 1998; Nolano et al., 1999). Both approaches have limited application to study human axonal regeneration due to pain with injection, variability with repeated applications and the long time needed to denervate the skin (weeks). The occlusive patch model addresses these limitations, producing a uniform lesion. Differences in regenerative capacity measured by this technique are not explained by variation in the degree of chemical axotomy, epidermal thickness or basement membrane impeding regenerating fibres. Rather, the rate of epidermal reinnervation appears to be a property of the individual studied.

The site of our biopsies was chosen after careful consideration. Peripheral neuropathy is a length‐dependent process making the lower limb, particularly the foot, more vulnerable, where a number of those with neuropathy have no ENFs. The thigh site increases the likelihood that a neuropathic population would have ENFs at baseline. The thigh is more conducive to the procurement of multiple biopsies than a constricted, more distal site such as the ankle. The application area was large enough, and biopsies were taken far enough from the treatment area margins to ensure that regenerative sprouting and not collateral sprouting was measured. Previous work has demonstrated that collateral sprouting occurs much more slowly and incompletely than regenerative sprouting (Theriault et al., 1998; Rajan et al., 2003). Therefore, it is highly unlikely that the regenerating epidermal nerve fibres observed after capsaicin application could represent collateral sprouts emanating from the capsaicin treatment border 5 mm away.

We rigorously addressed the possibility that capsaicin‐induced loss of ENF staining might not represent sensory axon loss, but rather loss of PGP 9.5 staining. Previous studies in control subjects have correlated loss of PGP 9.5 immunoreactivity with an increase in heat pain thresholds followed by return towards baseline values (Nolano et al., 1999). We extended these observations in several ways. First, we found that loss and regeneration of ENFs were associated specifically with changes in heat pain threshold, but not mechanical detection thresholds. Secondly, we observed equivalent results whether staining was performed with PGP 9.5 or Gα0, suggesting that loss of nerve fibres occurs and that the cytosolic antigenic target of PGP 9.5 (ubiquitin hydrolase) does not ‘leak out’ through stimulated capsaicin receptor 1 membrane channels (TRPV1). Lastly, EM studies from a biopsy taken after 97 days of regeneration provided structural evidence of denervation followed by regeneration.

In the studies with diabetic subjects, we found that the rate of ENF regeneration was markedly reduced compared with healthy control subjects. Subjects with diabetes had an average rate of ENF regeneration that was 42% that of control subjects, and this difference persisted even after adjusting for age, baseline ENF density and post‐capsaicin application ENF density. Although we only studied small nerve fibre morphology, there may be implications for large nerve fibre function/morphology. The time line of studies which depend on improving nerve function as assessed by large fibres, however, may be quite long.

Abnormalities in nerve regeneration in the setting of diabetes have been studied in animal models and inferred from analysis of human sural nerve biopsies. Reduced axonal growth has been observed in STZ (streptozotocin) diabetic rats compared with non‐diabetic rats following sciatic nerve crush (Bisby, 1980), or sciatic transection (Longo et al., 1986). Several potential mechanisms have been suggested for reduced regeneration in diabetic animal nerve, including ischaemia (Longo et al., 1986), abnormalities in macrophage activation (Kennedy and Zochodne, 2000), altered growth factor support (Tomlinson et al., 1997; Pierson et al., 2002) and impaired insulin action (Pierson et al., 2003). Similarly, pathological study of human sural nerve biopsies from people with diabetes suggests that while regeneration occurs, it is unable to compensate for fibre loss (Dyck et al., 1988; Bradley et al., 1995). Other studies have suggested that regenerative sprouting may be excessive in early diabetic neuropathy (Brown et al., 1976; Britland et al., 1990). Clinical studies of patients undergoing median nerve release at the wrist suggest that people with diabetes recover less well than those without diabetes (Ozkul et al., 2002). All of these studies have been qualitative in nature, inferring an abnormality in nerve regeneration in people with diabetes. Our results demonstrate impaired axonal regeneration in people with diabetes, and additionally quantify the regenerative deficit and provide a dynamic system by which to study nerve regeneration.

We investigated the role of different factors on regenerative capacity in both control subjects and those with diabetes. In both groups, age and gender were not associated with regenerative rate. The duration of diabetes, diabetes type and glycated haemoglobin level were also not significant factors among diabetic subjects. Duration of diabetes is an inherently inaccurate measurement, particularly in type II diabetes where the onset of disease is often not known. Similarly, a glycated haemoglobin value at study onset does not reflect glycaemic control over time. It may be that historic glycaemic control is important in determining the regenerative rate, as has been suggested for neuropathy progression in type I diabetes (Writing Team for the Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications Research Group, 2002). The presence of neuropathy was associated with a further reduction in the rate of regeneration, with neuropathy‐free diabetic subjects having a rate intermediate between that of controls and neuropathic subjects with diabetes.

Basement membrane changes are known to occur in diabetes (King et al., 1989), and these changes could potentially act as a barrier for regenerating axons growing through the dermis into the epidermis. If this were the case, then we would have expected to see fibres whose growth was arrested at the dermal side of the dermal–epidermal junction, which we did not. Furthermore, basement membrane thickening has been reported to be related to duration of diabetes and to male gender (Wendelschafer‐Crabb et al., 2003), neither of which were associated with regenerative rate in our studies.

The finding of impaired regeneration in diabetes has important implications for the clinical treatment of nerve injuries in the setting of diabetes. Treatment algorithms for evaluation and treatment of focal neuropathies such as median neuropathy at the wrist often involve a period of ‘watchful waiting’ or conservative treatment with steroid injection or splinting (Hughes, 2003). Given the results suggesting that a regenerative deficit exists in diabetes, other approaches may be appropriate in people with diabetes.

Our results have significant potential implications for DPN trial design. First, this model has the ability to measure regenerative capacity over a period of a few months, rather than years, which could reduce the length of regenerative DPN trials. This would remove an obstacle from existing trial designs as consensus has emerged that DPN trials should have longer durations, with several trials currently slated to have durations of several years (Pfeifer and Schumer, 1995; Pfeifer et al., 1997). Secondly, since this model identified a quantifiable nerve abnormality in diabetes without neuropathy, this population could be included in trials of promising regenerative agents.

We have described a standardized nerve injury model and studied the rate of regeneration in control and diabetic subjects with and without peripheral neuropathy. We found that people with diabetes have an impaired capacity to regenerate small calibre sensory nerve fibres and this deficit exists even among subjects without objective evidence of neuropathy. This suggests that abnormalities in nerve function begin early in diabetes, before signs or symptoms are apparent, and that regenerative capacity is a very sensitive measure of that dysfunction. This is consistent with the finding of significant axon loss being present in the setting of normal electrophysiology (Behse and Buchthal, 1978). Our results have implications for future regenerative DPN trials. Our findings provide a mechanism to efficiently evaluate promising regenerative agents in a matter of months, using relatively small sample sizes. Lastly, these findings offer a rationale by which to include non‐neuropathic subjects with diabetes in regenerative trials. Such a target study population is well suited to regeneration studies and may increase our ability to detect a treatment effect.


We wish to thank Dr Leland Scott for identifying Gα0 as an epidermal marker, Richard Skolasky for statistical expertise, David Cornblath for critical reading of the manuscript, and members of the Howard County, Pikesville, Anne Arundale and St Agnes Diabetes Support Groups for their help. This work was supported by grants NS41374 (M.P.), and RR00522 (GCRC JHU) from the National Institutes of Health, the Juvenile Diabetes Research Foundation (M.P.) and Vertex Pharmaceuticals Inc.


View Abstract