Design: Systematic review and meta-analysis of 153 study populations
Population: Over 132,000 peritoneal dialysis patients across 37 countries
Primary outcome: Peritonitis incidence (episodes per patient-year)
Effect: Global 0.31 (95% CI 0.28-0.34); regional range 0.21 to 0.99

Peritonitis is the complication that most often ends peritoneal dialysis, and knowing where a unit sits against global rates matters for quality improvement. This systematic review pooled 153 study populations, over 132,000 PD patients across 37 countries.

The global incidence was 0.31 episodes per patient-year, comfortably within the International Society for Peritoneal Dialysis benchmark of no more than 0.40, but the average hid wide variation. Regional rates ranged from 0.21 per patient-year in the Western Pacific to 0.99 in the African Region, and 10 of 37 countries exceeded the ISPD benchmark. Gram-positive organisms were the commonest cause, followed by culture-negative episodes.

The practical value is as a yardstick. A programme can compare its own peritonitis rate and organism mix against these contemporary regional figures, and a rate above 0.40 per patient-year signals a need to review connection technique, training and prophylaxis rather than accept it as inevitable.

Systematic review of over 132,000 peritoneal dialysis patients across 37 countries.
Global peritonitis incidence was 0.31 episodes per patient-year (95% CI 0.28-0.34).
Regional rates ranged from 0.21 (Western Pacific) to 0.99 (Africa) per patient-year.
10 of 37 countries exceeded the ISPD benchmark of 0.40 episodes per patient-year.
Gram-positive organisms were the commonest cause, then culture-negative episodes.

Why it matters

It provides a contemporary, regionally specific yardstick against which a PD programme can judge and improve its peritonitis rate.

The statistics, in plain English

A rate of 0.31 episodes per patient-year means roughly one peritonitis episode every three patient-years on average. The wide regional range (0.21 to 0.99) shows the global average should not be read as any single unit's expected rate.