Kidney Week

Abstract: FR-PO1119

Development and Validation of a Prediction Model to Estimate 24-Hour Urine Creatinine as a Muscle Mass Surrogate in Patients with Advanced CKD

Session Information

• Health Maintenance, Nutrition, and Metabolism
  November 07, 2025 | Location: Exhibit Hall, Convention Center
  Abstract Time: 10:00 AM - 12:00 PM

Category: Health Maintenance, Nutrition, and Metabolism

• 1500 Health Maintenance, Nutrition, and Metabolism

Authors

• Donnelly, Lauren, The University of Tennessee Health Science Center, Memphis, Tennessee, United States

Lauren Donnelly, MS

• Shrestha, Prabin, The University of Tennessee Health Science Center, Memphis, Tennessee, United States

Prabin Shrestha, PhD

• Sumida, Keiichi, The University of Tennessee Health Science Center, Memphis, Tennessee, United States

Keiichi Sumida, MD, PhD, MPH, FASN

• Kalantar-Zadeh, Kamyar, University of California Los Angeles, Los Angeles, California, United States

Kamyar Kalantar-Zadeh, MD, DrMed, PhD, MPH, FASN

• Kovesdy, Csaba P., The University of Tennessee Health Science Center, Memphis, Tennessee, United States

Csaba P. Kovesdy, MD, FASN

Background

Low muscle mass is associated with poor outcomes in patients with CKD, but measuring muscle mass is difficult. 24-hour urine creatinine (24hrUC) is a surrogate of muscle mass, but 24-hour urine collections are cumbersome. We aimed to develop a prediction model for 24hrUC for patients with non-dialysis dependent CKD.

Methods

We used logistic regression to develop a prediction model of low 24hrUC (defined as 24hrUC below median) using demographic, clinical, and laboratory data from a national cohort of 102,477 US Veterans who transitioned to dialysis from October 1, 2007, to March 31, 2015. We identified 3,008 patients with at least one 24hrUC within three years prior to initiating dialysis. We evaluated discrimination using Harrell’s C statistics and calibration by comparing predicted and observed values. We constructed a simplified model using variable importance analysis, and generated a risk score by transforming coefficients into integers. The model was internally validated using 70:30 data splitting.

Results

The mean (SD) age and eGFR were 64.6 years (10.5) and 26 ml/min/1.73m² (16), 62% were white and 97% were male. Low 24hrUC was associated with white race, unmarried status and presence of ischemic heart disease, cerebrovascular and peripheral vascular disease, and CHF. The complete model included 20 variables and had excellent discrimination (C statistic and 95%CI in the training and validation cohort: 0.72 [0.70 – 0.74] and 0.70 [0.67 – 0.74]) and calibration (Figure A). Variable importance analysis identified BMI, hypertension, sex, race, ethnicity, dementia, and age as the strongest predictors for our simplified model (C statistics: 0.68 [0.65 – 0.72] and 0.69 [0.66 – 0.73]) (Figure B).

Conclusion

We developed a prediction model from demographic and clinical features to accurately estimate low 24HrUC, a surrogate of sarcopenia, in patients with advanced CKD. Our model will need to be validated in different populations.

Funding

• NIDDK Support

Digital Object Identifier (DOI)

• doi: 10.1681/ASN.2025aawggd2y