Predictors of Variability in Urine Albumin-to-Creatinine Ratio in the General US Adult Population

Measured urine albumin-to-creatinine ratio (ACR) is known to be variable between measurements from the same individual with chronic kidney disease or diabetes, but the factors predisposing to increased variability remain unclear. Our objective was to quantify and identify predictors of short-term, within-person variability in albuminuria in a generalizable population.

This is a cross-sectional analysis of 4,829 adults in the 2009-2010 United States National Health and Nutrition Examination Survey who provided two urine samples within 10 days. We calculated the absolute percentage change, within-person coefficient of variation, and 95% reference change values for the overall population and within clinically meaningful subgroups. We then performed linear regression modeling to assess for a priori selected predictors of variability and examined nonlinear relationships using restricted cubic splines.

The mean age was 47.8 (SD 16.5) years, the mean eGFR was 95.8 (SD 20.4) mL/min per 1.73 m2, 11.3% had diabetes, and 51.5% of the participants were women. Factors associated with greater variability in temporally proximate repeat urine ACR samples included age, sex, and severity of initial urine ACR. Overall, albuminuria varied by approximately 28%, but there was heterogeneity within clinical subgroups. Reclassification into alternative albuminuria category following the second collection was low.

Patients' urine ACR can fluctuate roughly ±30% on average and confirmatory testing is wise in those with borderline values. Reviewing longitudinal trends in urine ACR or averaging several collections is likely more clinically meaningful than any single value.