Methodological issues in assessing the incidence of peritoneal dialysis-associated peritonitis in children.

OBJECTIVES: To evaluate the distribution of peritonitis incidence and assess the usefulness of patient-specific peritonitis rates in children.
DESIGN: 49 children on automated peritoneal dialysis (PD) followed during a 2-year observation period.
SETTING: Single-center, academic children's hospital. PATIENTS: 49 children aged 2 months to 18 years; 24 prevalent, 25 incident during the observation period. Cumulative observation time was 639 patient-months. MAIN OUTCOME MEASURES: Cohort-specific peritonitis incidence, median patient-specific peritonitis incidence, mean peritonitis incidence by gamma-Poisson (negative binomial) modeling, peritonitis-free survival by Kaplan-Meier life-table analysis.
RESULTS: 68 new peritonitis episodes and 21 relapses occurred in 27 patients. The distribution of patient-specific peritonitis incidence was bimodal, with a large group experiencing no or very few episodes, and another cluster around 1 episode per 6-9 months. Overall cohort-specific peritonitis incidence was 1.28, median subject-specific incidence 0.99, and mean incidence according to negative binomial modeling 1.04 (95% confidence interval 1.02-1.06) episodes per patient-year. Median peritonitis-free survival time was 6.9 months. In those patients who developed peritonitis, subject-specific peritonitis incidence was inversely correlated with patient age (r = -0.42, p < 0.05) and duration of chronic PD at last observation (r = -0.42, p < 0.05).
CONCLUSIONS: Since the distribution of peritonitis in children is non-Gaussian, the average risk of peritonitis is more accurately expressed by the median of the individual subject-specific peritonitis rates or by the mean incidence estimate obtained by the negative binomial distribution model. The assignment of a personal peritonitis risk to each patient permits risk factor analysis by routine statistical methods, even in smaller populations.