Cisplatin is a valuable antineoplastic drug which as a dose-limiting side-effect causes sensory neuropathy, and which therefore is often combined with less neurotoxic carboplatin. It has not been possible to reproduce cisplatin neuropathy in experimental animals, and the neurotoxic mechanism in man is disputed. We investigated post-mortem material from 12 patients and 15 control subjects. Half of the fibres with diameters of > or = 9 microns, or more than 15% of all fibres (P < 0.02), had disappeared in the sural nerves of patients. Signs of axonal regeneration were lacking. The dorsal root ganglia D12 and L2 of some but not of all patients contained necrotic neurons and nodules of Nageotte. The mean volume of the somata was reduced by 18% (P < 0.03). A relation between cumulated doses, treatment free interval and changes in nerve or ganglia was not found. The platinum content was high in all tissues except in the spinal cord when the patient had died shortly after treatment, and it decreased with increasing interval, least so in liver, sensory ganglia and sural nerves. The results support the hypothesis that cisplatin neuropathy is a neuroneopathy rather than a dying-back axonopathy.