This study evaluated the safety of triple vs single-lumen catheters in intravenous nutrition. Patients who were judged likely to benefit from a triple-lumen catheter were randomized to receive either a single-lumen catheter, with additional peripheral or central venous access as needed, or a triple-lumen catheter. All patients were at increased risk of catheter-related infection because of one or more of the following conditions: > 60 years of age, breakdown of skin integrity, severe underlying illness, diagnosis of acute pancreatitis, recent head or neck surgery, or presence of a preexisting infection. Patients were excluded who had neutropenia, were immunosuppressed, had body burns > 40%, or had contaminated wounds in the subclavicular area. Of 204 patients entered between June 1989 and November 1991, 177 completed the required > or = 7 days of therapy. Seventy-eight of these patients were randomized to a single-lumen catheter and 99 to a triple-lumen catheter. Catheters were inserted and maintained by the Nutrition Support Team. Dressings were monitored daily and changed weekly using a bio-occlusive dressing. When parameters were met for a possible septic episode, simultaneous peripheral and central catheter blood cultures were obtained using the Isolator method. Catheter-related sepsis was considered present if the colony count from a central catheter lumen was > or = 5 times that of the peripheral blood. The incidence of catheter-related sepsis for single-lumen catheters was 2.6% (2 of 78) compared with 13.1% for triple-lumen catheters (13 of 99) (p < .01). No correlation was found with the number of insertion attempts, catheter days, or patient's age.(ABSTRACT TRUNCATED AT 250 WORDS)