In patients with chronic kidney disease (CKD), high sodium intake has been associated with poorer blood pressure control and a greater risk of disease progression,1,2 whereas potassium intake appears to exert a protective effect at both the cardiovascular and renal levels.3 The assessment of the urinary excretion of these analytes is performed using 24-h urine collections; however, the collection process has significant limitations, with error rates that may range from 40 to 70% depending on the criteria applied.4 To minimize the effects of inadequate collections, anthropometry-based equations that estimate 24-h urinary creatinine excretion have been proposed, which would facilitate a reliable interpretation of the data even if the sample is not properly collected.5 Among the main validated equations for estimating creatinine excretion are those proposed by the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI),6 Kawasaki et al.,7 and Tanaka et al.,8 previously analyzed by our group in a retrospective study in which CKD-EPI showed the best overall performance.9 The aim of the present study was to prospectively evaluate the performance of the CKD-EPI, Kawasaki, and Tanaka equations for estimating urinary creatinine excretion, analyzing their correlation, precision, and agreement with creatinine measured in 24-h urine samples in patients with CKD.
We conducted a prospective observational study in incident patients referred to outpatient nephrology clinics for CKD evaluation. A total of 81 patients were included, from whom 149 serum samples, 142 spot urine samples, and 119 24-h urine determinations were obtained. Of the latter, 74 samples from 34 patients met the Imbembo and Walser criteria for adequate urine collection (creatinine excretion between 15–25 mg/kg in men and 10–20 mg/kg in women)10 and were considered valid for analysis. Although the study had a prospective design, the analysis was conceived as a per-determination validation, with statistical analysis performed per valid 24-h urine sample rather than per patient, since some individuals contributed more than one determination during the study period. The study was approved by the Research Ethics Committee of Hospital Universitario Infanta Leonor. All patients provided written informed consent.
Patients had a mean age of 64.7 ± 14.3 years; 80.2% had arterial hypertension and 40.7% had diabetes mellitus. A total of 82.6% were overweight or obese, with a median eGFR of 40.7 ml/min/1.73 m2. CKD stages G2–G4 predominated (G1: 2.8%; G2: 29.6%; G3: 33.8%; G4: 28.2%; G5nD: 5.6%), with nephroangiosclerosis being the most frequent etiology.
The CKD-EPI, Kawasaki, and Tanaka equations6–8 were used to estimate urinary creatinine. Correlation (r and R2), bias (defined as the mean of the differences between measured and estimated 24-h urinary creatinine), precision (P15, P30, and P50), and agreement by Bland-Altman analysis were evaluated. Comparative results and plots are presented in Table 1.
Correlations of the different creatinine estimates in 24-h urine from 74 spot urine samples.
| Equation | r | R² | Bias | 95% LoA (mg/day) | P15 (%) | P30 (%) | P50 (%) |
|---|---|---|---|---|---|---|---|
| CKD-EPI | 0.904 | 0.816 | −8.73 (193.72) | −399.96; +447.43 | 71.6 | 91.9 | 100 |
| Kawasaki | 0.458 | 0.210 | +120.77 (403.67) | −906.39; +828.13 | 29.7 | 71.6 | 87.8 |
| Tanaka | 0.856 | 0.733 | −163.53 (234.08) | −834.33; +443.84 | 52.7 | 82.4 | 93.2 |
r and R2: correlation coefficients between measured and equation-estimated creatinine (Cr) (Pearson, n = 74). Bias: mean absolute difference (mg/day) between measured and estimated 24-h urinary Cr. LoA: Bland-Altman limits of agreement (1.96 SD). Percentage of equation-estimated Cr values falling within 15, 30, or 50% of measured 24-h urinary Cr.
Values indicating the highest agreement or precision are highlighted in bold.
Despite prior patient training for 24-h urine collection, only 62.2% of samples (74/119) were valid. The CKD-EPI equation showed the best overall performance, with high correlation (r = 0.904; R2 = 0.816), minimal bias (−8.73 mg/day), and precision of 71.6, 91.9, and 100% for P15, P30, and P50, respectively. The Tanaka equation showed an acceptable correlation (r = 0.856; R2 = 0.733), although with greater bias (−163.53 mg/day), indicating a tendency to overestimate creatinine excretion relative to measured 24-h urinary creatinine. Conversely, the Kawasaki equation exhibited a lower correlation (r = 0.458; R2 = 0.210), greater dispersion, and lower precision (P15: 29.7%; P30: 71.6%). The Bland-Altman plots revealed narrower limits of agreement for CKD-EPI and Tanaka, whereas Kawasaki showed greater variability (Fig. 1).
Linear regression and Bland-Altman plots using the CKD-EPI, Kawasaki, and Tanaka equations in 74 urine samples meeting the Imbembo and Walser criteria.
Pearson correlation coefficients. The horizontal bars in the Bland-Altman model represent the mean ± 1.96 SD of the mean value determined from the difference between measured 24-h urinary Cr and Cr estimated by each equation.
In conclusion, the CKD-EPI and Tanaka equations represent useful and clinically applicable tools for estimating 24-h urinary creatinine excretion in patients with CKD. As a limitation, the reduced number of 24-h urine samples considered valid for analysis should be noted, which could limit the generalizability of results, as well as the use of some repeated determinations from the same individual, which may have influenced the correlation coefficients, bias, or limits of agreement.
In this context, the use of anthropometry-based equations, and specifically the CKD-EPI equation, to estimate 24-h urinary creatinine in CKD could be useful for more precisely estimating sodium, potassium, and protein excretion in 24-h urine samples (which is a test that is difficult for many patients to complete), and even for estimating 24-h urinary biochemistry using spot urine samples, although this potential application would require specific validation that was not performed in the present study.
