Development of a Multivariable Prediction Model for Citrate Accumulation in Liver Transplant Patients Undergoing...

Patients with impaired citrate metabolism may experience citrate accumulation (CA), which causes life-threatening metabolic acidosis and hypocalcemia. CA poses a challenge for clinicians when deciding on the use of regional citrate anticoagulation (RCA) for patients with liver dysfunction. This study aimed to develop a prediction model integrating multiple clinical variables to assess the risk of CA in liver transplant patients.

This single-center prospective cohort study included postoperative liver transplant patients who underwent continuous renal replacement therapy (CRRT) with RCA. The study end point was CA. A prediction model was developed using a generalized linear mixed-effect model based on the Akaike information criterion. The predictive values were assessed using the receiver operating characteristic curve and bootstrap resampling (times = 500) to estimate the area under the curve (AUC) and the corresponding 95% confidence interval (CI). A nomogram was used to visualize the model.

This study included 32 patients who underwent 133 CRRT sessions with RCA. CA occurred in 46 CRRT sessions. The model included lactate, norepinephrine >0.1 μg/kg/min, alanine aminotransferase, total bilirubin, and standard bicarbonate, which were tested before starting each CRRT session and body mass index, diabetes mellitus, and chronic kidney disease as predictors. The AUC of the model was 0.867 (95% CI 0.786-0.921), which was significantly higher than that of the single predictor (p < 0.05). A nomogram visualized the prediction model.

The prediction model integrating multiple clinical variables showed a good predictive value for CA. A nomogram visualized the model for easy application in clinical practice.