Besides the common distal symmetrical sensory-motor polyneuropathy (DSP) that is often associated with autonomic dysfunction, diabetic patients may develop multifocal sensory-motor deficits (MDN) secondary to roots, plexus and nerve trunk involvement. Nerve ischaemia has been suggested as a common mechanism for the different patterns of diabetic neuropathies, yet the important clinical differences that exist between DSP and MDN suggest concurrent factors. In order to learn more on the subject, we prospectively studied 22 consecutive diabetic patients with MDN, for which other causes of neuropathy were excluded by appropriate investigations, including biopsy of a recently affected sensory nerve. Three patients had a relapsing course, and the others an unremitting subacute-progressive course. Painful MDN progressed over 2-12 months. The neurological deficit predominated in distal lower limbs which were involved in all patients, unilaterally in seven, bilaterally in the others, with an asynchronous onset in most cases. In addition, a proximal deficit of the lower limbs was present on one side in seven patients, and on both sides in six. Thoracic radiculoneuropathy was present bilaterally in two patients, and unilaterally in one. The ulnar nerve was involved in one patient, and the radial nerve in two. The CSF protein ranged from 0.40 to 3.55 g/l; mean: 0.87 g/l. Electrophysiological testing showed severe, multifocal, axonal nerve lesions in all cases. Asymmetrical axonal lesions were found in all nerve specimens. The mean density of myelinated axons was reduced to 1340 +/- 1070 per mm(2) of endoneurial area versus 8370 +/- 706 myelinated fibres/mm(2) in controls. The mean density of unmyelinated fibres was reduced to 5095 +/- 6875 per mm(2) (extremes: 0-26 600). On teased fibre preparations, 34 +/- 31% of the fibres were at different stages of axonal degeneration (extremes 0-99%); 7 +/- 6% of the fibres showed segmental demyelination or remyelination. Necrotizing vasculitis of perineurial and endoneurial blood vessels were found in six patients. Endoneurial seepage of red cells was present in 11 specimens, and endoneurial haemorrhage in five. Ferric iron deposits that characterize previous bleeding were found in seven patients, including two who had no red cells in the endoneurium. Perivascular mononuclear cell infiltrates were present in the nerve specimens of 21 out of 22 patients, prominently in four patients. In comparison, nerve biopsy specimens of 30 patients with severe distal symmetrical diabetic polyneuropathy showed mild epineurial mononuclear cell infiltrate in one patient and endoneurial seepage of red cells in another. We conclude that MDN is related to pre-capillary blood vessel involvement in elderly diabetic patients with a secondary inflammatory response.