The potential of nighttime urine samples for epidemiologic research on sodium intake: Evidence from a population-based cohort study in Switzerland

Abstract

Background: The gold standard for measuring sodium intake is based on multiple 24-h urine sodium (UNa) collections, which are logistically complex and pose a high burden on study participants. Its major alternative, spot urine sampling, has been shown to lead to systematic bias in sodium intake estimation. Nighttime urine collections are a potential alternative, as they place a substantially lower burden on participants and are thus less likely to lead to underestimation of UNa compared with 24-h urinary collections. Objectives: We examined the possibility of estimating 24-h UNa excretion using nighttime collections only, both at the individual and population level. Methods: We used the data from the Swiss kidney project on genes in the hypertension cohort, containing 1757 24-h urine collections from 1090 individuals, divided into daytime and nighttime samples. The participants were adults of European ancestry living in Switzerland. We examined the ability to predict 24-h UNa excretion based on nighttime collections using the following 3 methods for sodium estimation: 1) normalization by collection duration, 2) normalization by mean urine volume, and 3) multivariable linear regression models. Results: The Pearson correlation coefficient for nighttime collections compared with the 24-h measurements were 0.661, 0.651, and 0.682–0.749 (depending on the regression model specifications) using the 3 nighttime collection methods of sodium estimation, respectively. All of the estimation methods led to different probability distributions from the target distribution. Method 1 showed systematic negative bias, method 1 and method 2 showed bias with respect to diabetes and hypertension status, and method 3 showed an overestimation of low excretion and an underestimation of high excretion. Conclusions: Estimating 24-h UNa excretion based on nighttime collections was found to be imprecise and/or biased at the individual level and failed to capture population-level characteristics beyond central tendency.