The natural history of intradural spinal cord arteriovenous shunts is unknown. We performed an observational study in a consecutive patient cohort with symptomatic intradural spinal cord arteriovenous shunts who were admitted to three institutes to investigate the clinical course of this complex disease, which would provide valuable evidence to inform clinical decision-making. The clinical course of patients with symptomatic intradural spinal cord arteriovenous shunts from initial presentation to occurrence of clinical deterioration, initiation of treatment, or last follow-up was analysed. Patients with at least 1 month of observation were included in this study. Clinical onset and deterioration patterns were divided into acute and gradual. Annual and cumulative rates of clinical deterioration as well as their risk factors were analysed using Kaplan-Meier life table analysis and Cox proportional hazards model. To assess risks and benefits of treatment, post-treatment clinical courses were further assessed. Four hundred and sixty-six patients with a mean observational period of 36.9 +/- 58.8 months were included; 56.7% of patients presented with acute onset, of whom 77.3% experienced spontaneous recovery. Age of onset older than 28 years, initial modified Aminoff and Logue scale of 43, mid-thoracic lesions and non-ventral lesions were independent predictors of failure for spontaneous recovery. The annual risk of general, acute and gradual clinical deterioration after onset was 30.7%, 9.9% and 17.7%, respectively. Risk of deterioration was highest in the early period after initial onset. Acute onset was the only independent risk factor [hazard ratio 1.957 (95% confidence interval, CI 1.324-2.894); P = 0.0008] of acute deterioration and gradual onset was the strongest predictor [hazard ratio 2.350 (95% CI 1.711-3.229); P < 0.0001] of the gradual deterioration among all the stratifying factors. After invasive treatment, complete obliteration was achieved in 37.9% of patients (138 of 364) and improved or stable clinical status was noted in 80.8% of patients. Forty-two patients (11.5%) experienced permanent complications. Overall post-treatment deterioration rate was 8.4%/year, and 5.3%/year if permanent complications were excluded. The natural history of symptomatic spinal cord arteriovenous shunts is poor, especially in the early period after onset, and early intervention is thus recommended. Initial onset pattern significantly affects the natural history of the lesion, which prompts a differentiated treatment strategy.