In recent years an impressive decrease in the incidence of CAPD-related peritonitis was observed in our centre, from 1.4 in the mid-eighties to 0.4 per patient year in 1991. In order to analyse which factors were most responsible for this decline, the present study was performed. From the start of our CAPD programme in 1982 until September 1991, 100 patients were enrolled. For each patient, time elapsed from catheter insertion until first peritonitis episode was recorded. Outcome was measured as the peritonitis-free interval in days. The following variables have been evaluated: age, gender, type of catheter, type of system, presence of diabetes mellitus, leakage, break-in period, presence of an exit-site infection, and performing surgeon. Data were analysed first by Kaplan-Meier product-limit estimate of survival (peritonitis-free interval). Thereafter Cox proportional hazard analysis was applied to the data, providing a conditional probability of peritonitis at each moment during follow-up, given a certain combination of risk factors. Our results show that the system, in conjunction with the type of catheter, was a decisive factor in the decline of the peritonitis rate in our centre. Patients on the twin-bag system (twin-bag group) showed a significant increase in the peritonitis-free interval in comparison with patients using other systems (non-twin bag group). Among the other variables analysed, only diabetes mellitus appeared to be relatively important. Episodes of culture negative peritonitis were more frequently observed in the twin-bag group, compared to the non-twin bag group. In absolute numbers Staph. non-aureus was the micro-organism most effectively reduced.