Assessing U.S. Sodium Intake through Dietary Data and Urine Biomarkers^,,

Abstract

Sodium intake is related to blood pressure, an established risk factor for heart disease and stroke. Reducing intake may save billions in United States health care dollars annually. Efforts targeting sodium reductions make accurate monitoring vital, yet limited information exists on the accuracy of the current data to assess sodium intake in the United States population. In this symposium, new findings were presented on the accuracy of estimating population 24-h urinary excretion of sodium from spot urine specimens or sodium intake from 24-h dietary recalls. Differences in accuracy by sex, BMI, and race were apparent as well as by timing of spot urine collections. Although some published equations appear promising for estimating group means, others are biased. Individual estimates of sodium intake were highly variable and adjustment for within-individual variation in intake is required for estimating population prevalence or percentiles. Estimates indicated United States sodium intake remains high.

Estimating 24-h Urinary Sodium Excretion from Casual (“Spot”) Urinary Sodium Concentration in Western Populations: INTERSALT Study

Paul Elliott provided an overview of the difficulty with monitoring population sodium intake (1), indicating 24-h urine collection was the recommended method of assessment. However, the expense and burden of 24-h urine collection can be prohibitive. Sodium excretion from spot urine samples, although less expensive and burdensome, has not been extensively evaluated for use in population surveys.

He presented results from a recent analysis of ∼6000 participants aged 20–59 y from 29 North American/European samples in the International Study of Salt, Other Factors, and Blood Pressure (INTERSALT) who collected a spot urine sample followed by a 24-h urine collection between 1985 and 1987. To guard against under- and overcollections of 24-h urine, the start and end times were recorded by staff at each clinic. Collections were rejected if participants reported losing “more than a few drops” of urine or if urine volume was <250 mL in 24 h. Participants were randomly assigned to test or validation datasets. Gender-specific regression equations were developed estimating individual 24-h urinary sodium excretion from spot urinary sodium, potassium, and creatinine concentrations, age, age², region, and BMI. Using the validation data, estimated 24-h sodium excretion with and without potassium in the models was compared with observed 24-h sodium excretion at the individual and population (sample) levels. Correlations were higher for population samples (r ∼0.7) than for individuals (r ∼0.5). Bias in estimated 24-h sodium excretion was small at the population level, but among individuals, overestimation of observed 24-h sodium excretion occurred at lower concentrations and underestimation at higher levels. Equations slightly overestimated the proportions of men and women with urinary 24-h sodium excretion above the recommended levels.

This is the first published analysis, to our knowledge, on use of spot urine samples for estimating group-level sodium intake using data from different population samples. Data were limited among important population subgroups, including blacks or African Americans and people with hypertension, and the timing of the sample was not analyzed. Despite these limitations, the results suggest casual urine samples may be a useful, low-burden, low-cost alternative to 24-h urine collections for estimating population sodium intakes. To increase validity in external populations, calibration with study-specific 24-h urinary collections is recommended.

A Calibration Study Assessing Sodium and Related Micronutrient Status from Timed-Spot and 24-h Urine Samples: Design and Initial Findings

Chia-Yih Wang presented design, methods, and descriptive results of a calibration study examining the use of timed-spot urine samples, as currently collected in NHANES, to estimate population-level intake of sodium, potassium, chloride, and iodine. Volunteers aged 18–39 y were recruited from the DC metropolitan area from June to August 2011; one-half were women and one-half were black/African American. Pregnant women and persons taking loop diuretics, with reported chronic kidney disease, or with new or modified hypertension treatment in the last 2 wk were excluded. From June to August 2011, participants collected urine at home. Urine collection began after discarding the first morning void and each void was collected in a separate container and ended with the first morning void the following day. Upon return of the urine, a 24-h dietary recall was collected. One-third of participants who provided an initial 24-h urine collection were asked to complete a second 24-h urine collection and dietary recall. A 24-h urine composite sample was prepared by taking an aliquot proportional to the volume of each void. In addition, 4 timed spot urine samples (morning, afternoon, evening, and overnight), corresponding to the times urine samples are collected in the NHANES, were selected from each 24-h collection. Completeness of urine collection was assessed during data collection by volume (≥500 mL), length of collection (≥20 h), and self-reported missing ≤1 void. Completeness was further verified after collection using the rate of observed to expected 24-h urine creatinine of ≥0.6. Approximately 85% of eligible volunteers participated and collected a complete sample (n = 407/481). Observed within-individual variation in sodium concentrations of spot samples illustrated the importance of accounting for this when developing equations to estimate population 24-h sodium excretion.

Validity of Published Equations Predicting 24-h Sodium Excretion in Adults Aged 18–39 y

Mary Cogswell, using data described by Dr. Wang, assessed the validity of previously published equations using a single, spot urine sodium concentration to estimate 24-h sodium excretion. The INTERSALT equation described by Dr. Elliott (1) and 3 published equations (2–4) were used to estimate 24-h sodium excretion from each of the 4 timed-spot urine samples. Across equations and timed samples, mean bias in estimated 24-h sodium excretion among all participants was highest using the Kawasaki et al. (2) equation (>1000 mg/d). Using morning, afternoon, or evening samples, bias was least with the INTERSALT equations. Using overnight samples, bias was least using the Tanaka et al. (3) or Mage et al. (4) equations but within race-sex subgroups was significant and in opposite directions. Individual correlations with observed 24-h sodium excretion were moderate (0.4–0.6) for all prediction equations and times of spot urine collection. Bland-Altman plots indicated significant over- and underestimation across low to high values of individual sodium intake. Although 24-h urine collection remains the recommended method for assessing population sodium intake, using spot urine INTERSALT equations may provide the least biased information about group mean sodium intake among young United States adults. Differences in the validity of estimated 24-h urine sodium excretion by timing of collection and race-ethnicity should be considered when applying prediction equations using spot urine samples to assess sodium intake.

Accuracy of the USDA Automated Multiple Pass Method for Assessing Population Sodium Intake

Donna Rhodes (5) presented analyses assessing the validity of the dietary intake instrument used in What We Eat in America, NHANES to collect sodium intakes compared with 24-h urine sodium excretion. Healthy volunteers aged 30–69 y were recruited from the DC metropolitan area; 465 completed both a 24-h dietary recall and a 24-h urine collection during the same period between July 2002 and June 2004. Reporting accuracy was calculated as the ratio of reported sodium intake from the dietary recall to that from the adjusted 24-h urine sample (24-h urinary sodium/0.86). Dietary data included salt added in cooking but not salt added at the table. More than 90% of participants had some college or greater education, 40% had a postgraduate degree, and 79% were non-Hispanic white. The mean reporting accuracy was >90% for all participants but differed by BMI, with the lowest accuracy among obese participants. Reporting accuracy varied by age among women, with higher accuracy at ages 50–69 y. Results suggest the USDA Automated Multiple Pass Method produces a valid measure for estimating sodium intakes at the group level but is less valid among overweight and obese adults. The generalizability of results to other age groups, those with less education, and by race-ethnicity deserves further investigation.

Estimated Change in Sodium Excretion in the United States between 1988 and 2010 Using Spot Urine Samples

Christine M. Pfeiffer noted the limited reports on temporal trends in sodium intake in the United States population. She presented preliminary sodium intake estimates from a single stored spot urine sample from NHANES 1988–1994 (convenience set), 2003–2006 (one-third random subset), and 2010 (one-third random subset). Sample selection for the first 2 time periods (n ∼1250 each) was stratified by blood pressure and sodium intake (24-h recall) to ensure a wide range of values for adults aged 20–59 y. About 600 urine samples from adults aged 20–59 y were analyzed in 2010. Estimated 24-h urine sodium excretion was calculated using the INTERSALT equations (1). Statistical sample weights were used to generate national estimates. Overall, sodium intake estimated from both dietary data and stored spot urine samples increased slightly during ∼20 y. Estimated mean 24-h urine sodium excretion differed by sex, BMI, and blood pressure, whereas mean sodium intake from 24-h recalls differed by sex but not BMI or blood pressure. Estimates of average sodium intake were in excess of guidelines.

Conclusions

In this symposium, new data were presented on the accuracy of estimating population sodium intake from a single spot urine sample and 24-h dietary recalls. Spot urine samples and 24-h dietary recalls have promise for monitoring temporal trends in average United States sodium intake. Although the 24-h urine collection is used as the gold standard, bias related to low participation and completeness in 24-h collections should be weighed against better acceptability but potentially lower accuracy of spot samples. Differences in accuracy by sex, BMI, race-ethnicity, and diurnal variation in sodium excretion (spot urine samples) need to be accounted for when estimating United States sodium intake. Although methods appear promising for estimating group means, individual estimates were variable and adjustments for within-individual variation in intake or excretion are required to estimate population distributions.

Ongoing national efforts to better monitor United States sodium intake also include testing equations using more than one spot urine sample to estimate population distributions of sodium intake, analyzing data from a NIH-funded urine calibration study among older adults (∼50% with hypertension), piloting 24-h urine collection in NHANES, and updating sodium and other nutrients in the USDA nutrient databases to minimize error in dietary data. Regardless of the assessment method, current estimates indicate average United States sodium intake remains high.