Abstract
BACKGROUND: Adjusting dietary potassium intake based on 24-hour urinary potassium excretion is the primary method of preventing hyperkalemia. Currently, there is no accurate and convenient method for calculating maximum 24-hour urinary potassium excretion in kidney failure without replacement therapy patients. We developed and validated two new models to assess the upper limit of dietary potassium consumption in this high-risk cohort, using the maximum 24-hour urinary potassium excretion as a proxy. METHODS: The data of 145 kidney failure without replacement therapy patients with hyperkalemia was gathered. The prediction models were developed using multilayer perceptron and stepwise multiple linear regression utilizing a stochastic sample of 102 (70%) patients. Within the rest 43 (30%), the performance of various models was independently verified. RESULTS: The two new models had low bias (-0.02 and -0.57 mmol/24h vs 66.74 and 79.91 mmol/24h, mean absolute error = 5.57 and 5.22 vs 68.95 and 81.37), high accuracy (percentage of calculated values within_±30% of measured values = 83.45% and 84.14% vs 0.00% and 0.00%), high correlation with measured values (Spearman correlation coefficient = 0.72 and 0.72 vs 0.46 and 0.45, intraclass correlation coefficient = 0.67 and 0.70 vs 0.03 and 0.03) and high agreement with 24-hour urine potassium measurements (95% limits of agreement of Bland-Altman plot = 13.70 and 13.20 mmol/24h vs 113.8 and 191.3 mmol/24h). CONCLUSION: These new models show high clinical application value for the calculation of maximum 24-hour urinary potassium excretion in kidney failure without replacement therapy patients with hyperkalemia.