Between 1980 and 1984, of 107 patients receiving 16 mg/d of dexamethasone for spinal cord compression, three (2.8%) developed gastrointestinal (GI) perforation and two (1.9%) GI bleeding; of 226 being tapered from 100 mg/d of dexamethasone, perforation occurred in six (2.7%) and GI bleeding in eight (3.5%). Of 125 patients with GI perforations treated between 1979 and 1986, 41 (33%) were on steroids, 24 for neurologic disease. Median duration of steroid therapy was 24 days; 20 (91%) of the neurologic patients perforated within 30 days. The steroid group had more free peritoneal involvement (p less than 0.00001), but fewer signs and symptoms of peritonitis (p less than 0.000001) than the nonsteroid group. Seventeen patients were receiving steroids for cord compression; they had significantly more rectosigmoid perforations (p less than 0.014) and associated constipation (p less than 0.000001) than the 108 remaining patients. GI perforation is a less well-recognized complication of steroid therapy in neurologic patients than is GI bleeding though it occurs as frequently, is more difficult to diagnose, and far more serious. In steroid-treated patients, prevention of constipation might avert this serious complication, while early diagnosis will improve the outcome.