One hundred one continuous ambulatory peritoneal dialysis (CAPD) patients from a single North American center were analyzed in a retrospective and cross-over study for peritonitis rates using a standard system (Travenol System II) or a Y-shaped disconnect-disinfectant system (Travenol O-set). Twenty-one of 34 patients using the standard set (group I) had 53 episodes of peritonitis in 508 patient-months or one episode per 9.6 patient-months. Nine of 17 patients switching from the standard to the disconnect-disinfectant system (group II) experienced 22 episodes of peritonitis in 275 patient-months or one episode per 12.5 patient-months on the standard set, while six patients had 10 episodes of peritonitis in 275 patient-months or one episode per 27.5 patient-months on the disconnect-disinfectant system (P less than 0.04). Twenty-eight of 67 new CAPD patients starting on the disconnect-disinfectant system (group III) had 37 episodes of peritonitis in 1,086 patient-months or one episode per 29.4 patient-months (P less than 0.01 v group I). Exit-site infections (ESI) occurred in 35.3% of patients using the standard set versus 34.3% of those using the O-set. The presence of an ESI was not associated with a higher risk of peritonitis, but modified the bacteriological profile of subsequent peritonitis episodes in patients using the O-set, favoring the organisms isolated from the exit site. Decreases in peritonitis rates with the O-set were due to a reduction of peritonitis episodes secondary to most bacterial agents and not only to skin organisms. Diabetics using intraperitoneal insulin had similar peritonitis and ESI rates as nondiabetics.(ABSTRACT TRUNCATED AT 250 WORDS)