We thank Deriaz et al.1 for their proposal to use a regression model including sodium, potassium, and an interaction term between the two rather than sodium-to-potassium ratio, to determine whether or not this interaction is significant. We repeated our analysis in the CKD-REIN cohort,2 using a linear model that included spot urinary sodium-to-creatinine (uNa/Cr), potassium-to-creatinine (uK/Cr), and an interaction term to estimate the effect on systolic blood pressure, while adjusting for potential confounders. In Table 1, we show the adjusted mean difference in systolic blood pressure expressed in mm Hg by quartiles of uNa/Cr for a median uK/Cr, by quartiles of uK/Cr for a median uNa/Cr, and by quartiles combining increasing uNa/Cr quartiles with decreasing uK/Cr quartiles to estimate joint effects.
Table 1.
Adjusted variations in systolic blood pressure by quartiles of spot urine sodium-to-creatinine and potassium-to-creatinine ratios
| Mean difference in systolic blood pressure (mm Hg) | |
|---|---|
| Spot uNa/Cr (mmol/g) at median uK/Cr | |
| Q1 <69.6 | Ref |
| Q2 (69.6–102.3) | 1.31 (0.79 to 1.83) |
| Q3 (102.3–142.7) | 2.61 (1.59 to 3.63) |
| Q4 ≥142.7 | 5.12 (3.15 to 7.10) |
| Spot uK/Cr (mmol/g) at median uNa/Cr | |
| Q1 <36.0 | Ref |
| Q2 (36.0–46.9) | −0.28 (−0.91 to 0.34) |
| Q3 (46.9–61.1) | −0.53 (−1.70 to 0.64) |
| Q4 ≥61.1 | −0.98 (−3.16 to 1.20) |
| Joint effects of spot uNa/Cr and uK/Cr | |
| Q1 of uNa/Cr and Q4 of uK/Cr | Ref |
| Q2 of uNa/Cr and Q3 of uK/Cr | 1.59 (0.31 to 2.87) |
| Q3 of uNa/Cr and Q2 of uK/Cr | 3.09 (1.01 to 5.16) |
| Q4 of uNa/Cr and Q1 of uK/Cr | 5.89 (2.67 to 9.13) |
uK/Cr, urine potassium-to-creatinine ratio; uNa/Cr, urine sodium-to-creatinine ratio; Ref, reference.
Model adjusted for age, gender, education level, estimated glomerular filtration rate, albuminuria category, history of diabetes, heart failure, dyslipidemia, body mass index, and number of antihypertensive drug classes.
As in our primary analysis,2 systolic blood pressure significantly increased across spot uNa/Cr quartiles (P = 0.003), up to 5.12 (3.15 to 7.10) mm Hg between the fourth (Q4) and the first quartile (Q1), while it decreased, although nonsignificantly, across spot uK/Cr quartiles (P = 0.256). The interaction between spot uNa/Cr and uK/Cr was not statistically significant (P = 0.121). As a result, the size of the joint effect was modest as reflected by a mean difference of 5.89 (2.67–9.13) mm Hg in systolic blood pressure between Q4 (combining high spot uNa/Cr and low uK/Cr) and Q1 (low uNa/Cr and high uK/Cr), just above that observed for spot uNa/Cr alone. However, we would caution against a risk of bias of such a model due to colinearity between the 2 urinary markers (rho = 0.30). As shown by Yoo et al.,3 even moderate correlations between 2 covariates may impact regression estimates.
Our conclusions about the preponderant role of sodium in blood pressure control, and the marginal, if any, role of potassium, in patients with moderate to severe chronic kidney disease seem robust to model specification.
References
- 1.Deriaz D., Maes N., Kolh P., Krzesinski J.-M. Sodium and potassium intake: a new statistical model to test their effects on health outcomes. Kidney Int Rep. 2020;5:2403. doi: 10.1016/j.ekir.2020.09.017. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Alencar de Pinho N., Kaboré J., Laville M. Urinary sodium-to-potassium ratio and blood pressure in CKD. Kidney Int Rep. 2020;5:1240–1250. doi: 10.1016/j.ekir.2020.05.025. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Yoo W., Mayberry R., Bae S. A study of effects of multicollinearity in the multivariable analysis. Int J Appl Sci Technol. 2014;4:9–19. [PMC free article] [PubMed] [Google Scholar]