We undertook this study to test the hypothesis that perioperative total parenteral nutrition (TPN) decreases the incidence of serious complications after major abdominal or thoracic surgical procedures in malnourished patients. We studied 395 malnourished patients (99 percent of them male) who required laparotomy or noncardiac thoracotomy. They were randomly assigned to receive either TPN for 7 to 15 days before surgery and 3 days afterward (the TPN group) or no perioperative TPN (the control group). The patients were monitored for complications for 90 days after surgery. The rates of major complications during the first 30 days after surgery in the two groups were similar (TPN group, 25.5 percent; control group, 24.6 percent), as were the overall 90-day mortality rates (13.4 percent and 10.5 percent, respectively). There were more infectious complications in the TPN group than in the controls (14.1 vs. 6.4 percent; P = 0.01; relative risk, 2.20; 95 percent confidence interval, 1.19 to 4.05), but slightly more noninfectious complications in the control group (16.7 vs. 22.2 percent; P = 0.20; relative risk, 0.75; 95 percent confidence interval, 0.50 to 1.13). The increased rate of infections was confined to patients categorized as either borderline or mildly malnourished, according to Subjective Global Assessment or an objective nutritional assessment, and these patients had no demonstrable benefit from TPN. In contrast, severely malnourished patients who received TPN had fewer noninfectious complications than controls (5 vs. 43 percent; P = 0.03; relative risk, 0.12; 95 percent confidence interval, 0.02 to 0.91), with no concomitant increase in infectious complications. The use of preoperative TPN should be limited to patients who are severely malnourished unless there are other specific indications.