We report our experience with 24 consecutively treated patients (15 men and 9 women, median age 64 years) with anti-myelin-associated glycoprotein (anti-MAG) neuropathy. The rates of response to plasma exchange (40%), immune globulin (16%), and cyclophosphamide-based therapy (36%) were similar. Five (24%) responded to the first treatment modality, 32% to a second, alternative modality, and 31% to a third. Only 4 of 12 responders had sustained improvement; the others relapsed after a median of 7 months. In those 4 patients, the median immunoglobulin M (IgM) level dropped by 25% compared to an increase of 24% in the nonresponders (P = 0.04). Thus, most patients with anti-MAG neuropathy failed to have sustained improvement after treatment, and none of the therapies emerged as superior. Disability improved transiently after therapy in approximately 50% of cases. A 25% reduction of the IgM level predicted sustained improvement, but was difficult to achieve. There were no clinical or electrodiagnostic features associated with a treatment response, nor did a reduction of the anti-MAG antibody titer correlate with clinical improvement.