Longitudinal CBC-Derived Biomarkers and Peritonitis in Peritoneal Dialysis: A Prentice, Williams and Peterson Recurrent-Event Analysis

Background: Peritonitis remains a major complication of peritoneal dialysis (PD), and most previous studies have focused little on the recurrent burden of peritonitis. We investigated the associations of longitudinal complete blood count (CBC)-derived biomarkers with first and subsequent qualifying peritonitis episodes using a recurrent-event framework. Methods: Incident PD patients enrolled between 28 December 2011 and 3 November 2021 were followed through 31 December 2022. Each CBC-derived biomarker was modeled separately as a time-varying exposure in an event-order-stratified Prentice, Williams and Peterson total time (PWP-TT) model of 30-day-separated qualifying peritonitis episodes, adjusted for confounding factors. Restricted cubic splines (RCS) revealed the true shape of the relationship. Biomarkers were assessed separately with Benjamini–Hochberg false discovery rate (BH-FDR) correction for multiple comparisons. Sensitivity analyses included a common complete-case analysis, PWP gap time (PWP-GT), and patient-level shared frailty models. Results: A total of 374 incident PD patients, 10,057 lab records and 179 peritonitis episodes were included in the analysis. Higher neutrophil-to-lymphocyte ratio (NLR) was associated with increased risk of qualifying peritonitis episodes in the multivariate PWP-TT analysis (HR 1.073, 95% CI 1.025–1.123, BH-FDR-adjusted p = 0.018). RCS analysis showed a nonlinear association for NLR (p overall < 0.001, p nonlinearity = 0.022), with risk increasing at higher NLR levels. Systemic inflammation response index (SIRI) showed a similar association (HR 1.099, 95% CI 1.021–1.184, BH-FDR-adjusted p = 0.050) with no evidence of nonlinearity (p nonlinearity = 0.112). Findings were unchanged in the common complete-case analysis and directionally consistent in PWP-GT and shared frailty models. Conclusions: Higher NLR was robustly associated with an increased risk of first and subsequent qualifying peritonitis episodes, with a nonlinear pattern. Whereas SIRI showed a similar but less robust association. These findings support the potential utility of NLR for longitudinal peritonitis risk assessment in PD.