CKD

CKD predicts one-year post-nephrectomy chronic kidney disease (CKD) risk by integrating early postoperative serum creatinine measurements with preoperative and intraoperative factors in renal cell cancer patients.

Key Features:

• Predictive modeling: Incorporates variables collected preoperatively, on postoperative days 0–5 and at one month, emphasizing serum creatinine trajectory to predict CKD one year after radical nephrectomy (RN) or partial nephrectomy (PN).
• Optimal sampling time point: Identifies postoperative day 4 (POD 4) serum creatinine as the optimal single sampling time point for predicting one-year CKD stages.
• Meta-modeling: Combines various subsets of developed models to generate 120 meta-models intended to enhance prediction accuracy and reliability.
• Prognostic variables: Considers independent prognostic factors including etiology, hemoglobin level, creatinine level, proteinuria, and urinary protein/creatinine ratio.
• Validation performance: Demonstrates high area under the curve (AUC) values with excellent calibration and discrimination across 1-, 2-, and 3-year predictions in training and validation sets.

Scientific Applications:

• Clinical decision support: Identifies patients at higher risk for CKD progression after nephrectomy to inform timely interventions and management planning.
• Research utility: Provides a resource for studying the impact of nephrectomy on renal function and factors influencing CKD development.

Methodology:

Models were developed from retrospective data of 1,556 patients who underwent laparoscopic or robotic RN or PN using univariate and multiple Cox proportional hazards analyses to identify independent prognostic factors; various model subsets were combined to produce 120 meta-models, and performance was validated by AUC, calibration, and discrimination for 1-, 2-, and 3-year predictions in training and validation sets.