Does a predictive model based on urine conductivity improve the screening performance for suspected early-stage CKD compared to urine protein in adult patients?
Predictive models based on urine conductivity provide a low-cost, rapid, and highly sensitive screening tool for early-stage CKD compared to urine protein.
(1) Background: Chronic kidney disease (CKD) affects more than 800 million global population. Early detection followed by clinical management is among the best approaches for the affected individuals. However, a sensitive screening tool is not yet available. (2) Methods: We retrospectively reviewed 600 patients aged >20 years with a full range of estimated glomerular filtration rate (eGFR) for clinical assessment of kidney function between 1 January 2020, to 30 April 2021, at the Taichung Veterans General Hospital, Taichung, Taiwan. With stratified sampling based on the level of eGFR, participants were evenly grouped into training and validation sets for predictive modeling. Concurrent records of laboratory data from urine samples were used as inputs to the model. (3) Results: The predictive model proposed two formulae based on urine conductivity for detecting suspected early-stage CKD. One formula, Pₘale45, was for used male subjects aged ≥45 years, and it had a prediction accuracy of 76. 3% and a sensitivity of 97. 3%. The other formula, Pfemale55, was used for female subjects aged ≥55 years. It had a prediction accuracy of 81. 9% and a sensitivity of 98. 4%. Urine conductivity, however, had low associations with urine glucose and urine protein levels. (4) Conclusion: The two predictive models were low-cost and provided rapid detection. Compared to urine protein, these models had a better screening performance for suspected early-stage CKD. It may also be applied for monitoring CKD in patients with progressing diabetes mellitus.
Wu et al. (Fri,) studied this question.