Research letters
Spinal-cord involvement in diabetic peripheral neuropathy

Simon E M Eaton, Nigel D Harris, Satyan M Rajbhandari, Pam Greenwood, Iain D Wilkinson, John D Ward, Paul D Griffiths, Solomon Tesfaye

The pathogenesis of diabetic distal symmetrical polyneuropathy (DSP) is poorly understood but there is some evidence that the disease process might extend beyond peripheral nerves. We used magnetic-resonance imaging to measure spinal-cord cross-sectional area in diabetic patients with and without DSP and in healthy controls. There were significant differences in cord area between the groups at C4/5 and T3/4 (p=0004 and p=0033, respectively), with a smaller cord area in those with DSP compared with controls (p=0001 and p=0016 for C4/5 and T3/4, respectively). These results indicate that DSP is not simply a disease of the peripheral nerve and that there is substantial involvement of the spinal cord.

Lancet 2001; 358: 35-36

Distal symmetrical polyneuropathy (DSP) is a common complication of diabetes mellitus but the pathogenesis remains poorly understood1 and there is no proven rational treatment. Most of the clinical research into DSP has concentrated on functional and structural aspects of the peripheral nerve2 whilst other potentially important areas, such as the spinal cord, have been overlooked. There is, however, limited evidence from necropsy2,3 and electrophysiological studies,4 that the nerve damage in DSP may be more generalised, and might involve the spinal cord. In addition, the observation that electrical spinal-cord stimulation is ineffective in relieving neuropathic pain in patients with severe loss of vibration and joint-position sense, has led to the suggestion that the spinal cord may be involved in the disease process.5 We examined the involvement of the spinal cord in DSP by measuring cord area using magnetic-resonance (MR) imaging.

19 diabetic patients with DSP were compared to ten without DSP (non-DSP) and ten healthy controls (table). All participants were men and the groups were matched for age and height. The diabetic groups were matched for duration and control of diabetes. Exclusion criteria included substantial back problems, or vascular disease and neuropathy not caused by diabetes. The neurological status of all participants was assessed including sural and peroneal nerve-conduction velocities (NCV), vibration-perception threshold (VPT) on the great toe of the right foot, and cooling-detection threshold on the dorsum of the left foot. MR imaging was done using a standard spinal phased-array receive-only radiofrequency coil on a system operating at 15 Tesla (Eclipse, Marconi Medical Systems, Cleveland, OH, USA). Sagittal imaging of the entire spinal cord was done to exclude compressive lesions. T2 weighted scans were then done axially at three anatomical levels (level with the intervertebral discs at C4/5, T3/4, and T9/10) using a two-dimensional gradient echo technique (TE=179 ms; TR=800 ms; *=40; slice thickness=4 mm; in-plane resolution=078x096 mm). Spinal-cord cross-sectional area was measured at each level by a computer-assisted manual outline technique. Ten measurements were taken for each level and averaged with the assessor blinded to the name and group of the patient. Intraobserver and interobserver coefficients of variation were 25% and 3%, respectively. The scan-rescan coefficient of variation, measured after re-imaging seven randomly selected individuals after repositioning them in the scanner, was 38%. Data were compared by ANOVA.

  Controls Non DSP DSP
  (n=10) (n=10) (n=19)
Age (years) 490 (113) 496 (54) 508 (101)
Height (cm) 176 (58) 173 (76) 176 (54)
Weight (kg) 80 (74) 78 (109) 91 (130)
Types 1 diabetes/Type 2 diabetes   7/3 9/10
Duration of diabetes (years)   156 (113) 149 (86)
HbA1c (%)   84 (13) 84 (23)
Glucose (mmol/L)   113 (46) 125 (35)
Vibration-perception threshold (volts) 79 (16) 72 (17) 365 (140)
Sural nerve-conduction velocity (m/s) 489 (30) 481 (58) 278 (94)
Peroneal nerve-conduction velocity (m/s) 446 (30) 433 (36) 314 (76)
Values are mean (SD).
Clinical characteristics of patients

Significant differences in mean cord area were found at the levels of C4/5 (998 [SD 87] vs 929 [107] vs 869 [85] mm2 for controls vs non-DSP vs DSP respectively; p=0004;) and T3/4 (573 [33] vs 565 [67] vs 519 [61] mm2; p=0033, figure) but not T9/10 (550 [59] vs 542 [31] vs 530 [72] mm2; p=0689). Individual group comparisons showed that mean cord area was significantly lower in the DSP group compared with controls at both C4/5 (p=0001) and T3/4 (p=0016). Cord area in the non-DSP group was intermediate to, but not significantly different from the other groups, although at T3/4 the difference between the DSP and non-DSP groups was approaching significance (p=0074). These findings are not due to potential baseline differences between the groups as there were no significant correlations between cord area and any of the measured demographic or neurophysiological parameters. There was a difference in weight between the groups, with the DSP group averaging approximately 10 kg more than the other two groups, but it is unlikely that this will explain the differences we have found in cord area.

Spinal-cord area at C4/5 and T3/4 levels

Cord level C4/5: controls vs DSP, p=00001; non-DSP vs DSP, p=0114. Cord level T3/4: controls vs DSP, p=0016; non-DSP vs DSP, p=0074. Bars indicate mean values. DSP=diabetic symmetrical polyneuropathy.

Whether spinal-cord involvement is a primary or secondary event in DSP is not clear. One possibility is that damage to the peripheral nerve causes secondary spinal-cord shrinkage and causes the cord to die back. It is also possible that the primary damage is to the central nervous system, with the peripheral changes occurring secondarily. However, the necropsy findings of microvascular disease within the spinal cord,2,3 similar to that seen in the peripheral nerve, would suggest that the same pathogenetic mechanisms are involved in both areas and that the processes are concomitant.

This research is a pilot study with small numbers of participants, so there is a limit to the conclusions that can be drawn. Further work is required to improve the sensitivity of the measures and prove conclusively that the changes we found are due to a neuropathic process rather than simply the diabetes itself. Additionally, relating the cord area to the severity of neuropathy may help to understand the time course of spinal-cord involvement. Our observation of a small spinal-cord area in DSP suggests an important new direction for further research concentrating on detecting the neuropathic process at an earlier stage during which potential new therapies would be more likely to succeed.

1 Ward JD, Tesfaye S. Pathogenesis of diabetic neuropathy. In: Pickup J, Williams G, eds. Textbook of diabetes. 2nd edn. Oxford: Blackwell Science, 1997: 49.1-49.19.

2 Reske-Neilsen E, Lundbaek K, Rafaelsen OJ. Pathological changes in the central and peripheral nervous system of young long-term diabetics: II. Spinal cord and peripheral nerve.  Diabetologia 1968; 4: 34-43. [PubMed]

3 Slager UT. Diabetic myelopathy.  Arch Pathol Lab Med 1978; 102: 467-69. [PubMed]

4 Zeigler D, Muhlem H, Dannehl K, Arnold Gries F. Tibial nerve somatosensory evoked potentials at various stages of peripheral neuropathy in insulin dependent diabetes mellitus.  J Neurol Neurosurg Psychiatry 1992; 56: 58-64. [PubMed]

5 Tesfaye S, Watt J, Benbow SJ, Pang KA, Miles J, MacFarlane IA. Electrical spinal-cord stimulation for painful diabetic peripheral neuropathy.  Lancet 1996; 348: 1698-701. [Text]

Diabetes Research Unit (S E M Eaton MRCP, S M Rajbhandari MRCP, J D Ward MD, S Tesfaye MD) and Medical Physics (N D Harris PhD), Royal Hallamshire Hospital; and Section of Academic Radiology, University of Sheffield (P Greenwood, I D Wilkinson PhD, P D Griffiths MD), Sheffield, UK

Correspondence to: Dr Solomon Tesfaye, Royal Hallamshire Hospital, Glossop Rd, Sheffield S10 2JF, UK (


NDC Neurology Editor's Review:-

Diabetic peripheral neuropathy is common and frequently difficult to treat. The precise mechanism is unclear; postulated pathologies range from vasculitis and infarction to the metabolic consequences of glycemic dysregulation upon neuronal and supporting structures. In this study, Eaton et al have measured the cross-sectional area of three regions of the spinal cord including the cervical and lumbar expansions. In patients with a distal symmetrical polyneuropathy (DSP) the spinal cord cross sectional area was significantly reduced in 2 of the 3 regions compared with controls. Diabetics with and without clinical evidence of a neuropathy however did not differ significantly. The authors investigated intraobserver, interobserver and within patient measurement variability; the differences were outside these coefficients of variation for 2 of the 3 sites. It was concluded from this small pilot study, that these preliminary data were consistent with a contributory spinal cord pathology in DSP, although it could not be determined whether this was primary or secondary to the peripheral process. The contribution of these findings to symptomatology and treatment response is as yet unclear, especially as diabetics without DSP also seem to have some degree of spinal cord abnormality. Likewise, without pathological or possibly spectroscopic investigation of the spinal cord, it is unclear whether integrity is disturbed or whether this merely represents generalized volume change due to an effect on interstitial or intracellular fluid. -Dr. Stephen Oppenheimer

Professor of Neuroscience, Neurology and Medicine
Johns Hopkins University and NJ Neuroscience Institute

Senior Medical Director, Neuroscience
PharmaNet, Inc.

Neurology Branch,
The National Diabetes Center

RELOAD National Diabetes Center FRAMES