sW u

formed for suspected spinal dural arteriovenous fistulas on other imaging studies. The imaging and clinical data were reviewed in all patients who had or were suspected of having a spinal dural arteriovenous fistula and who had a spinal MR (n 5 44) and a myelogram (n 5 37). RESULTS: Spinal dural arteriovenous fistulas were more common in males (3.4:1) with an average age of 62 years (range, 37 to 81 years). The average time from onset of symptoms to diagnosis was 27 months. Clinical findings included weakness (55%), a progressive clinical course (100%), and a myelopathy on exam (84%). The nidus of the fistula was located between T-6 and T-12 in 61%, in the sacrum in 9%, and intracranially in 8%. In the spinal dural arteriovenous fistula group, vessels were seen on supine myelography in all patients. MR findings in this group included increased T2 signal in the cord (100%), gadolinium enhancement (88%), mass effect (45%), and flow voids (T1, 35%; T2, 45%). The patients in the negative spinal angiogram group were younger (average age, 51 years), had symptoms longer (average time from symptom onset to spinal angiogram, 59 months), and presented with numbness or pain (76%). When compared with the patients with spinal dural arteriovenous fistula, acute or stable deficits were more common (31%), and myelop-athy on exam was less common (56%). Although the angiogram-negative patients commonly had vessels on the myelogram (92%), abnormal T2 signal in the cord was unusual (17%). CONCLUSIONS: In the appropriate clinical setting, high T2 signal of the spinal cord is the most sensitive imaging finding in spinal dural arteriovenous fistula. The presence of mass effect and enhancement should not discourage this diagnosis. The likelihood of finding a spinal dural arte-riovenous fistula in a patient without T2 signal on MR is low