Nomogram for predicting selective renal arterial embolization in post-PCNL hemorrhage among patients with chronic kidney disease.

PURPOSE: Percutaneous nephrolithotomy (PCNL) is the standard treatment for large or complex renal calculi; however, postoperative hemorrhage requiring selective renal arterial embolization (SRAE) remains a rare but serious complication. Chronic kidney disease (CKD) may further increase bleeding risk, but predictive models for SRAE in this population are lacking. To develop and validate a nomogram for predicting the need for SRAE in patients with CKD who experience hemorrhage after PCNL.

METHODS

This retrospective cohort study included 3153 patients who underwent PCNL in two tertiary hospitals between June 2010 and June 2025. Of these, 986 had CKD. A total of 331 patients with post-PCNL hemorrhage and CKD were analyzed, including 299 who received conservative management and 32 who underwent SRAE. Clinical variables were compared between groups, and independent risk factors were identified using multivariate logistic regression. A nomogram was developed from the significant predictors and validated using receiver operating characteristic (ROC) analysis, calibration curves, and decision curve analysis (DCA).

RESULTS

Among 331 patients with post-PCNL hemorrhage and CKD, 32 (9.7%) required SRAE. Compared with the conservative group, SRAE patients had a higher rate of prior ipsilateral kidney intervention (31.3% vs. 11.4%), more acute hemorrhage (40.6% vs. 14.4%), and greater hemoglobin drop (39.5 ± 10.7 g/L vs. 25.3 ± 8.8 g/L), and a notably higher proportion of the absence of hydronephrosis (all p < 0.05). Multivariate analysis identified hydronephrosis grade, past ipsilateral intervention, hemorrhage type, and hemoglobin drop as independent predictors of SRAE (p < 0.05). The nomogram incorporating these variables achieved excellent discrimination (AUC = 0.901, C-index = 0.897) and good calibration (Hosmer-Lemeshow, χ² = 6.357, p = 0.607), with robust performance in the external validation cohort (AUC = 0.893). Decision curve analysis demonstrated favorable clinical utility. Among all 3,153 PCNL cases, patients with CKD had a significantly higher incidence of hemorrhage (33.6% vs. 19.3%, p < 0.001), while the SRAE rate was comparable to those with normal renal function (3.2% vs. 2.8%, p = 0.508).

CONCLUSION

The nomogram developed in this study provides a reliable and individualized tool for predicting the need for embolization among CKD patients who have already developed post-PCNL hemorrhage. Incorporating variables such as hemoglobin drop, hemorrhage type, hydronephrosis grade, and prior ipsilateral intervention, this model can assist clinicians in early risk stratification and optimize decision-making for timely intervention.