Increased morbidity after operation has been associated with long-term steroid therapy. To determine the correlation between steroid therapy and such morbidity, the perioperative course of 55 steroid-treated patients was reviewed: 27 had bronchopulmonary disorders (group P) and 28 had non-pulmonary diseases (group NP). There were six (11 per cent) deaths, of which three were steroid related. Among the 13 non-lethal postoperative complications, eight were considered to be steroid related in group P and one in group NP. The duration of steroid therapy was for a median of 24 months (range 1-408 months) in group P and for a median of 6 months (range 1-240 months) in group NP (P less than 0.01). In contrast, the daily dose of hydrocortisone or equivalent before operation was significantly lower in group P, with a median of 0.51 mg kg-1 day-1 (range 0.20-2.56 mg kg-1 day-1) than in group NP, with a median of 1.20 mg kg-1 day-1 (range 0.23-7.38 mg kg-1 day-1) (P less than 0.01). In conclusion, bronchopulmonary disorders requiring a long duration of steroid therapy are associated with a higher risk of steroid-related complications after surgery. A convenient mathematical model is proposed which may allow a preoperative assessment of surgical risk, using steroid dose and duration of treatment.