Recognized risk factors for metrizamide myelography are seizure disorder, seizure-threshold-lowering drugs, dehydration, and possibly age. After observing serious neurologic complications in diabetic patients after routine metrizamide myelography, a retrospective study was conducted to determine if diabetes should be considered another independent and important risk factor. Forty-one diabetic patients who had lumbar metrizamide myelograms were compared with a control group of 110 nondiabetic patients. A significantly higher incidence was found of severe vomiting (15% vs. 3%, p less than 0.01) and neurologic complications (20% vs. 2%, p less than 0.001) in the diabetic population. Neurologic complications included one case each of seizure, severe encephalopathy, auditory and visual hallucinations, and prolonged somnolence and four cases of confusion-anxiety. Four of the diabetic patients had major transient elevations of blood pressure. These findings suggest that diabetics are a high-risk population for metrizamide myelography. The dose of metrizamide should be minimized, whenever possible. The new nonionic myelographic agents may prove to be safer in this population, but caution and careful follow-up should be exercised in the initial trials with these patients.