PNI, CONUT, and GNRI for predicting treatment failure in peritoneal dialysis-associated peritonitis: a two-center...

Peritoneal dialysis-associated peritonitis (PDAP) is one of the common causes of technique failure in patients undergoing peritoneal dialysis. It may also affect long-term survival. Early identification of patients with a high risk of treatment failure is therefore important in clinical practice. Malnutrition has been reported to be related to poor outcomes in PDAP, but the predictive value of different nutritional scores has not been fully compared.

This retrospective study included patients diagnosed with PDAP at two centers between January 2015 and December 2025. Clinical data at admission were collected. Nutritional status was evaluated by the Prognostic Nutritional Index (PNI), the Controlling Nutritional Status (CONUT) score, and the Geriatric Nutritional Risk Index (GNRI). Receiver operating characteristic (ROC) curves and area under the curve (AUC) were used to compare the predictive ability of these three scores. Restricted cubic splines (RCS) were used to observe the dose-response relationship. Multivariable logistic regression was used to analyze the risk factors for treatment failure. Subgroup analyses were further performed to test the stability of the results. Net reclassification improvement (NRI) and integrated discrimination improvement (IDI) were used to evaluate the additional predictive value of the nutritional scores.

A total of 393 patients with PDAP were included. The rate of treatment failure was 30.79%. Among the three nutritional scores, PNI had the highest predictive value, with an AUC of 0.712. The AUCs of CONUT and GNRI were 0.681 and 0.633. All three nutritional scores showed linear dose-response relationships with treatment failure (p for nonlinearity: PNI = 0.242, CONUT = 0.748, GNRI = 0.952) and were identified as independent predictors in multivariable-adjusted analyses (all p < 0.001). Incremental analysis showed that adding PNI significantly improved the predictive performance over the baseline model (AUC: 0.627 to 0.723, p < 0.001), while CONUT showed moderate improvement (AUC: 0.627 to 0.700, p = 0.003) and GNRI did not (AUC: 0.627 to 0.662, p = 0.091). NRI and IDI confirmed the superior incremental value of PNI. Subgroup analyses showed no significant interactions (all p > 0.05).

PNI demonstrated superior predictive and incremental value for treatment failure in PDAP compared with CONUT and GNRI, suggesting that it may serve as a practical tool for early risk stratification in this population.