Work

Predictors of mortality and technique failure in incident peritoneal dialysis patients: a competing-risks analysis

Year: 2026

Type: article

Abstract: In peritoneal dialysis (PD), PD discontinuation is commonly used to benchmark program performance. However, death may account for a larger share of treatment attrition. We examined predictors of PD discontinuation and mortality and evaluated a simplified cardiovascular-kidney-metabolic (CKM) composite score for mortality risk stratification. We performed a retrospective cohort study of 134 incident PD patients at a Malaysian tertiary center (2023–2025). PD discontinuation was defined as permanent transfer to haemodialysis for PD-related complications. Cause-specific Cox models and Fine–Gray subdistribution models were used to assess predictors of PD discontinuation and on-PD mortality. A CKM composite score incorporating diabetes mellitus, hypertension, cardiovascular disease (CVD), dyslipidaemia, and hypoalbuminaemia was evaluated against a three-variable model (cardiovascular disease, diabetes, serum albumin) for mortality discrimination using the time-dependent area under the curve (AUC). Over 166 person-years (median follow-up 14 months), 34 attrition events occurred: 25 deaths and 9 PD discontinuations. PD discontinuation was uncommon and its recorded causes were predominantly PD-related infective and mechanical complications (refractory peritonitis in 5 of 9). Peritonitis burden and PD-related hospitalisation were not modelled as predictors of discontinuation: peritonitis episodes accrue during follow-up and could not be dated, the precipitating episode is itself counted in the exposure, and admission criteria overlap the discontinuation endpoint. On-PD mortality was associated with baseline cardiometabolic-nutritional burden (cause-specific HR 1.77 per point, 95% CI 1.22–2.58, p = 0.003; Fine–Gray sHR 1.79, 95% bootstrap CI 1.20–3.43, p = 0.035). Apparent discrimination was similar for the CKM composite score and the three-variable model (time-dependent AUC at 2 years 0.81 vs. 0.79; event-weighted mean AUC 0.68 vs. 0.69, difference not distinguishable from zero); because the models share most covariates this reflects overlapping information rather than added prognostic value. In this exploratory single-center cohort, death accounted for a larger proportion of observed PD attrition than PD discontinuation. Baseline cardiometabolic-nutritional burden was associated with on-PD mortality, whereas PD-related infective and mechanical complications predominated among the recorded reasons for PD discontinuation. The CKM composite score showed apparent mortality discrimination similar to a three-variable model; as both share most components, external validation is required before any clinical use. Not applicable. This is an observational cohort study and was not a clinical trial.

Source: BMC Nephrology

DOI: 10.1186/s12882-026-05395-x