Validation of spot urine in estimating 24-h urinary sodium, potassium and sodium-to-potassium ratio during three different sodium diets in healthy adults

Abstract Purpose To evaluate the validity of spot urine assay methods in estimating the 24-h urinary sodium, potassium and sodium-to-potassium ratio during three different sodium diets. Materials and methods Twelve healthy volunteers were asked to adhere to 3 dietary sodium targets (3.3–5.0g/day,<3.3 g/day and >5.0 g/day) for three consecutive weeks and to measure salt excretion daily in spot urine samples using a self-monitoring device. On day 7 of each week, 24-h urine was collected to compare measured with estimated 24-h salt excretion (by the Kawasaki, Tanaka and INTERSALT equations). Results Correlation coefficients relating measured and estimated 24-h sodium excretion were low and not significant for Kawasaki and INTERSALT and moderate for the Tanaka equation (τ 0.56–0.64,p<.05). Bland–Altman plots showed considerable differences between estimated and measured sodium excretion across all salt diets. Over 40% of the participants showed an absolute difference between measured and estimated 24-h sodium of more than 1000 mg/day. The correlation coefficients between 24-h and spot Na/K ratio were 0.67, 0.94 and 0.85(p<.05), and mean differences were 0.59, 0.06 and 0.48 for the intermediate, low and high sodium diets, respectively. Conclusion These findings do not support estimation of individual 24-h salt excretion from spot urine by the Kawasaki, Tanaka, or INTERSALT formula. Plain language summary Accurate monitoring of salt intake is essential to improve BP control. At present, measurement of sodium and potassium excretion in multiple non-consecutive 24-h urinary collections is considered the gold standard for measuring dietary sodium intake. However, this method is burdensome, time-consuming and error prone. Therefore, we assessed and compared the validity of three formula-based approaches to estimate 24-h urinary sodium and potassium excretion and the Na/K ratio from spot urine samples measured by a self-monitoring device under three different sodium diets using 24-h urine collections as the reference. We conclude that use of three commonly used equations that estimate 24-h urinary sodium and potassium excretion result in substantial bias, poor precision and poor accuracy and are therefore not recommended. The Na/K ratio based on multiple casual urine samples may be a useful, low-burden, low-cost alternative method to 24-h urine collection for monitoring daily salt intake.


Introduction
High dietary salt intake is a major contributor to the onset and progression of hypertension [1], a leading modifiable risk factor for cardiovascular disease (CVD) and mortality worldwide [2].Of the global number of deaths due to CVD, approximately 1.65 million per year have been attributed to excessive salt intake [3].As a consequence, the reduction in sodium consumption to levels below 5 g per day has been encouraged by the World Health Organisation (WHO) and various international treatment guidelines [4][5][6].However, despite rigorous governmental campaigns to reduce salt intake, a large gap remains between the recommended and actual daily intake of salt [7].In addition to the fact that conventional population-based approaches have failed to close this gap [8], individuals are still poorly aware of their daily salt intake, with individuals who reported that they were on a low-salt diet actually showing salt intake levels similar to those who were not on a lowsalt diet [9].
Self-monitoring devices that give individuals immediate feedback on their salt intake could provide support in achieving the goals as set by the WHO.At present, measurement of sodium excretion in multiple non-consecutive 24-h urinary collections is considered the gold standard for measuring dietary sodium intake [10].However, this method is burdensome, time-consuming and error prone [11].Hence, several formulas for estimating 24-h urinary sodium excretion from casual spot urines have been developed [12][13][14].Although this formula-based approach offers a lower patient burden, the most commonly used formulas for estimating the 24-h urine sodium excretions showed poor agreement with measured 24-h sodium excretion [15].In addition, these formulas require information on parameters such as body mass index and creatinine levels, which complicates the purpose of self-monitoring.To overcome these issues, the sodium-to-potassium ratio (Na/K ratio) has been proposed as an easier and potentially more reliable alternative for self-monitoring [16][17][18].However, few studies evaluated the Na/K ratio and the formula-based approach in parallel [19].Therefore, the aim of the current study was to assess and compare the validity of both the estimated 24-h urinary sodium and potassium excretion and the Na/K ratio from spot urine samples measured by a self-monitoring device under three different sodium diets using 24-h urine collections as a reference standard.

Study participants
Between June 2018 and October 2018, 12 healthy volunteers from the general population who expressed interest during the initial research and development phase were approached and prospectively enrolled in the University Medical Centre Utrecht (UMCU), the Netherlands.All participants were 18 years or older and motivated to adhere to the study protocol.Eligible participants were apparently healthy not having a known medical history of cardiovascular disease, chronic kidney disease, hypertension (blood pressure >140 and/or >90 mmHg at screening), pregnancy, incontinence, or an impaired vision.Moreover, subjects that were prescribed diuretics were not eligible for inclusion in the study.The study was approved by Medical Ethics Committee Utrecht (approval number 18-002/D) and conforms to the 1964 Declaration of Helsinki ethical guidelines.Participants were given oral and written information regarding the purpose and outline of the study, and written informed consent was obtained from each individual.

Dietary protocol
All participants started with their intermediate salt diet (3.3-5.0 g sodium per day), subsequently followed by a low salt diet (< 3.3 g sodium per day) and finished with a high salt diet (> 5.0 g sodium per day).Each dietary period lasted for 7 days.This particular order of the diets was chosen for two reasons.First, since the intermediate salt diet was expected to be easier to follow than the other two diets, it was planned first such that the participants could get used to the protocol for collecting and measuring urine samples.Second, because it takes several days for renal sodium and potassium excretion to adjust to a new salt diet [20][21][22], the high sodium diet was planned last to minimise its influence on sodium excretion during the other two diets.To support participants in reaching the weekly sodium intake target, all participants received oral and written dietary recommendations compiled by an experienced dietician.Participants received no dietary advice regarding potassium intake.Besides dietary restrictions, no other behavioural rules were prescribed to participants.

Spot urine collection
The urine collection schedule is shown in Table 1.Participants were instructed to collect daily spot urine samples and measure sodium, potassium and creatinine concentration by using the validated Medimate self-monitoring device (Fisic BV, Enschede, the Netherlands) (Supplementary Table 1) [23,24].Urine had to be collected in a pre-filled plastic container in which the urine sample was mixed with both an internal standard and buffer solution.By applying a drop of this mixed urine onto the disposable cartridge (using a pipette), sodium, potassium and creatinine concentrations were measured using microchip capillary electrophoresis (lCE) combined with conductivity detection.The device displayed the results within a few minutes and then stored them in the device's internal memory.On day 1, 3, 5 and 6, participants had to collect a morning spot urine sample in fasting state.On day 2 and 4, this had to be a random urine sample collected in a non-fasting state.On day 7, four spot urine samples (single morning fasting urine, before lunch (09:00 -13:00), before diner (13:00-17:00) and after dinner (17:00-23:00)) had to be collected.

Spot urine measurement by the medimate selfmonitoring device
The Medimate 2017 is a point-of-care self-test device for whole blood and urine.This device is able to measure lithium, sodium, potassium, magnesium, creatinine, chloride, phosphate and lactate in blood and urine using lab-on-a-chip technology, also referred to as 'Micro Total Analysis Systems' (lTAS).Measurements are performed using microchip capillary electrophoresis (lCE) combined with conductivity detection.The device consists of two main components: a multireader and a disposable cartridge (Supplementary Figure 1).The multireader contains all of the measurement electronics including a programmable black and white display on top and buttons to help the user perform a self-test, navigate through user-settings and scroll through stored measurements.The power connector is 100-240 V and at has a build in mini USB connector.The disposable cartridge called 'the Medimate Lab-Chip' (Supplementary Figure 2) consists of three parts: a glass chip (LoC) for measurement analysis (Supplementary Figure 3), a plastic housing and a seal.Measurements are performed by applying a drop of urine onto the disposable cartridge and inserting it into the multireader.Measurement time is less than 8 min.

Estimation of 24-h urine sodium and potassium excretion
The methods for the estimation of 24-h urinary sodium excretion by using spot urine samples were the Kawasaki method [12], the Tanaka method [13] and the INTERSALT method [14].The INTERSALT formula was not designed to estimate urinary potassium excretion, and therefore, only the Kawasaki and Tanaka method were used to estimate 24-h urinary potassium excretion.The estimation formulas used are listed in Supplementary Table 2.In accordance with the original design, for the Kawasaki formula, the estimations of 24-h sodium and potassium were based on the concentration of sodium, potassium and creatinine of the morning spot urine sample taken on the day following the 24-h urine collection, and for Tanaka and INTERSALT formulas, this was based on the average of spot urine samples 2, 3 and 4 taken on the same day as the 24-h urine collection (Table 1).

Twenty four-hour urine collection
Twenty-four-hour urine was collected on day 7 of each week.To ensure a complete urine collection, written and oral instructions were provided to the participants.Participants were asked to discard their first voided urine after getting up and to collect all the urine they voided during the subsequent 24 h, including the first urine of the following morning.In addition, participants were asked to record the start and finish times of the collections and report any wasting of urine to the study team.After completion of the 24-h urine collection, a vacuum test tube was filled with urine from the collection container and sent to the UMCU chemical laboratory.Twenty-fourhour urine collection was considered to be successful if the participant reported collecting urine !24 h, the total urine volume was greater than 1000 ml, and the difference between the expected 24-h creatinine excretion (based on sex, age and weight [25]) was within 25% of the actual measured values of 24-h urinary creatinine excretion.Urine collections were considered to be unsuccessful when the participant reported that portions of urine were missing from the collection.Unsuccessful samples were excluded from the analysis.Urinary creatinine, sodium and potassium values were determined by the ion-selective electrode method.All specimen measurements were conducted on a Beckman Coulter analyser (DxAU 5811) (Brea, CA, USA).

Statistical analysis
The differences between urinary sodium and potassium excretion measured by 24-h urine collection and that estimated by each of the 3 formulas were calculated.Data were presented as means ± SD.Paired ttests were used to evaluate the differences between the estimated and measured 24-h values for significance.The correlation between estimated and measured 24-h urinary sodium and potassium excretion was evaluated by the Kendall rank correlation coefficient.Bland-Altman analysis was also applied to provide a visualised assessment of the agreement between the measured and estimated 24-h salt excretion values.In addition, we analysed the absolute and relative differences between the estimated and measured values, where the absolute differences were calculated as (estimated valuemeasured value) and the relative difference as [(estimated valuemeasured value)/measured value Â 100%].The proportional distribution of the relative and absolute differences provided a graphical representation of the accuracy at the individual level.The relative difference in this study was defined in five groups: within 10%, 10% to 19%, 20% to 29%, 30% to 39% and over 40%.The absolute difference was also divided into five groups that were within 500, 500 to 1000, 1000 to 1500, 1500 to 2000 and over 2000 mg/day in sodium or potassium amount.Furthermore, we categorised measured 24-h sodium and potassium excretion into tertiles and then compared for each estimation method the fraction of participants misclassified into the wrong group.
Lastly, all of the above analyses were repeated to compare the average Na/K ratio measured in multiple casual spot urine samples on the 7 th day with the Na/K ratio measured in the 24-h urine collections.
All analyses were performed with R statistical software (Version 3.5.1;R foundation for Statistical Computing, Vienna, Austria).All p values were twotailed, with statistical significance set at .05.

Characteristics study population
Characteristics of the 12 participants included in the study are presented in Table 2.The mean age of subjects was 47 (interquartile range (IQR) 35-69) years, 11 subjects (92%) were male.Median body mass index (BMI) was 24 kg/m 2 (IQR 23-27) and median blood pressure was 130/77 mmHg (IQR 125-133/74-81).Figures 1 and 2 demonstrate the overall trend in daily spot sodium and Na/K ratio levels, respectively.Spot sodium and Na/K ratio levels changed in accordance with the sodium diet applied; low values during the low sodium diet and high values during the high sodium diet.Spot potassium levels remained stable over time (Supplementary Figure 4).
The number of successful 24-h urine collections was 10, 11 and 12 for week 1, week 2 and week 3, respectively.The mean ± standard deviation (SD) of the measured 24-h sodium excretion was 4011 ± 1369 mg in the intermediate sodium week, 1292 ± 562 mg in the low sodium week and 5488 ± 1493 mg in the high sodium week (Table 3).Mean measured 24-h potassium excretion was 3376 ± 987 mg in the intermediate sodium week, 3557 ± 719 mg in the low sodium week and 3844 ± 1128 mg in the high sodium week (Supplementary Table 3).

Measured versus estimated 24-h urine sodium excretion
The differences between estimated and measured 24-h urinary sodium excretion during each sodium diet are displayed in Table 3.During the intermediate sodium diet, the INTERSALT formula had the least mean difference between estimated and measured 24-h sodium excretion (À61 mg/day, 95% confidence interval (95%CI) À1310-1188).During the low sodium diet, all formulas consistently overestimated 24-h sodium excretion with the Tanaka formula having the lowest  mean bias (754 mg/day, 95%CI 586-922) and during the high sodium diet, all formulas underestimated the 24-h sodium excretion with the Kawasaki formula showing the least mean bias (À340 mg/day, 95%CI À1234-553).For potassium, both the Kawasaki and Tanaka formula consistently underestimated the 24-h excretion (Supplementary Table 3).

Relationships between the measured and estimated 24-h urinary sodium and potassium excretion
The Kendall rank correlation coefficients for the relationships between the measured values and the estimated 24-h sodium excretion values using the Kawasaki and INTERSALT methods were weak and not significant (Figure 3(A,C)).The estimated 24-h sodium excretion by the Tanaka method showed reasonable correlation with the measured 24-h sodium (s 0.56-0.64,p < .05)(Figure 3(B)).For potassium, similar results were obtained for the Kawasaki method.However, the correlation coefficient between measured and estimated 24-h potassium excretion by the Tanaka formula during the intermediate and high sodium diet were 1.0 (p < .05)and 0.56 (p < .05),respectively (Supplementary Figure 5).

Agreement between the measured and estimated 24-h urinary sodium and potassium excretion
The Bland-Altman plots analysis showed that the mean differences between estimated and measured 24-h sodium excretion for the three formulas had great variations across the different salt diets (Figure 4(A-C)).For potassium, the Bland-Altman plots showed that both the Kawasaki and Tanaka formula underestimated the 24-h potassium excretion and that variation in mean differences remained stable over the three diets (Supplementary Figure 6(A,B)).

Misclassification analysis
The degree of misclassification by the formula-based approach was assessed by comparing the tertiles of measured sodium and potassium intake to the tertiles of estimated sodium and potassium intake by the three estimation equations.(Table 4).The results showed that the percentage of individuals that were misclassified when using the Kawasaki method was 27%, and those for the Tanaka and INTERSALT methods were 24% and 30%, respectively.The overall underestimation of potassium by the Kawasaki and Tanaka formula resulted in a misclassification rate of 67% and 64%, respectively (Supplementary Table 4).

Spot sodium-to-potassium ratio versus 24-h sodium-to-potassium ratio
The overall median value of Na/K ratio in 24-h urine was 1.2 (IQR 1.0-1.4)during the intermediate sodium diet, 0.4 (0.3-0.4) during the low sodium diet and 1.4 (1.1-1.9)during the high sodium diet (Table 5).The average Na/K ratio of 3 casual spot urines taken on the same day as the 24-h urine collection was higher than 24-h urine Na/K ratios during the intermediate and high sodium diet and similar during the low sodium diet (Table 5).Note: All data in mean ± standard deviation.a Using the first morning urine sample collected on the day after the 24-h urine collection.b Using the average of casual urine spot 2, 3 and 4 collected on the same day as 24-h urine collection.
Relationships between spot Na/K ratio and 24-h Na/K ratio Correlation analysis on Na/K ratios from spot urine indicated high positive correlations with 24-h Na/K ratios (Figure 3(D)).Kendall rank correlation coefficients between the spot Na/K ratio and 24-h Na/K ratio were 0.67, 0.94 and 0.85 during the intermediate, low and high sodium diets, respectively (p values all <.05).
Agreement between spot Na/K ratio and 24-h Na/K ratio Using the Bland-Altman method, the bias between the 24-h urine Na/K ratio and the spot urinary Na/K Misclassification analysis Na/K ratio Misclassification analysis showed that over all three diets, 66% of the participants were classified into the matching 24-h Na/K ratio categories with the spot urine Na/K ratio (Table 6).

Discussion
In the present study, in which apparently healthy people were subjected to three different sodium diets, we found that three commonly used equations (Kawasaki, Tanaka and INTERSALT) performed poorly in estimating 24-h urinary sodium excretion from spot urine using a self-monitoring device.For all three sodium diets, estimated sodium and potassium excretion from each of the three equations showed low correlation with measured values.Moreover, all equations showed wide limits of agreement and relatively high rates of misclassification of salt intake groups.In contrast, the Na/K ratio based on multiple casual spot urines showed high correlation and acceptable agreement with 24-h Na/K ratio, especially during the low sodium diet.Collection of at least three non-consecutive 24-h urine samples is considered the most accurate method for estimating salt intake [10].However, this method is cumbersome and imposes a large burden on the patient.Therefore, a simple and valid method for estimating 24-h urinary sodium and potassium excretion from casual spot urine samples would be desirable.However, this study, investigating three simple formulas to estimate sodium and potassium excretion, found poor individual correlations of estimated and measured 24-h urinary sodium and potassium excretion, which is widely supported by previous studies in various populations [19,[26][27][28][29][30][31].For example, results of previous studies showed correlation coefficients for the relationships between the values estimated using the Kawasaki, Tanaka and INTERSALT formula of at most 0.54 in European populations [15,31], 0.43 in Chinese populations [19,28,32,33] and 0.18 in a South African population [30].
The Bland-Altman analysis in this study showed that the degree of under-and overestimation of all three formulas depended on the level of salt intake (¼ differential bias) which is in line with observations in previous studies [34][35][36].The INTERSALT equation appeared to overestimate at low levels, underestimate at high levels and estimate correctly at intermediate levels of individual sodium intake.These findings are consistent with results from a study Table 5. Measured 24-h sodium-to-potassium ratio compared to spot sodium-to-potassium ratio.conducted in young American adults [34] but are inconsistent with findings from the PURE (Prospective Urban Rural Epidemiology) China Study which showed that the Kawasaki formula yielded the lowest bias [27].This may be explained by the fact that the Kawasaki formula was developed in a Japanese population, whereas the INTERSALT formula was developed in European and American populations, both with their own dietary patterns and behaviours.
To further analyse the bias of sodium excretion estimates at the individual level we calculated relative and absolute differences between estimated and measured 24-h urinary sodium excretion.These analyses showed that the proportion of both relative and absolute differences above a certain amount was quite large for all three equations.More than 40% of the participants showed an absolute difference between measured and estimated 24-h sodium excretion, regardless of the equation used, of more than 1000 mg/day (¼ 2.5 g salt per day).When participants were further classified into tertiles of measured 24-h sodium excretion, almost 30% of the participants were misclassified.These findings are in line with previous studies [28,32,37,38] and further illustrate the potential for large errors in the estimation of individual 24-h urinary sodium excretion.
Taken together, our findings suggest that the commonly used formula-based approach for estimating 24-h sodium and potassium excretion produces a high degree of variability that could lead to misclassification of salt intake at the individual level and thereby to erroneous associations between salt intake and cardiovascular risk [39,40].To support (self-) monitoring and evaluation of salt reduction, more accurate methods for evaluating individual 24-h urinary sodium and potassium excretion are needed.As shown in our study, the Na/K ratio based on multiple casual urine samples might be a plausible alternative.We found that the correlation coefficients for the 24h Na/K ratio and spot Na/K ratio were 0.67, 0.94 and 0.85 and mean biases were 0.59, 0.06 and 0.48 for the intermediate, low and high sodium diets, respectively.This relatively good performance of the spot Na/K ratio in comparison with estimated sodium and potassium excretion has been described previously [16][17][18][19] and can probably be explained by the ratio being independent of urine volume and creatinine excretion, of which the latter may degrade if samples are not kept at proper temperature [41].This makes estimation of the Na/K ratio from casual urine easier than estimating sodium or potassium excretion alone.
Since the Na/K ratio has also been shown to be a superior measure than either sodium or potassium alone in relation to blood pressure, incident hypertension and CVD [42,43], this appears to be a preferable method for self-monitoring of salt intake that may motivate individuals to reduce their sodium intake, increase their potassium intake and thereby potentially improve blood pressure control and cardiovascular prognosis.At present, there is no generally accepted recommendation for the desired Na/K ratio.According to WHO reports, achieving the guidelines for both the sodium (<2 g/day) and potassium (>3.5 g/day) intake would result in a Na/K ratio of less than 1.00.
Strengths of this study include the collection of urine samples over a relatively long period covering different sodium diets enabling validation of the three equations and Na/K ratio over different levels of sodium excretion.Also, use of repeated casual urine samples for the calculation of the Na/K ratio most likely increased precision of the individual measurement of Na/K ratio and is in line with previous recommendations regarding use of the Na/K ratio for estimating individual salt intake [17,18].Finally, to provide more insight into the accuracy of all methods at the individual level, we not only evaluated the validity of the three equations by analysing the mean difference and correlation coefficients but also assessed the distribution of relative and absolute differences, as well as the misclassification of salt intake groups.
A limitation of this study is the study population being limited to relatively young and apparently healthy men.Therefore, it is unknown whether our findings also apply to individuals with other characteristics such as women, older individuals and patients with various disease, e.g.chronic kidney disease, atherosclerotic diseases or diabetes.Second, we applied the Kawasaki equation, originally developed for use with second morning urine voids, to spot urine samples collected in the morning since the protocol did not require it to be second morning urine voids.Thus, by not specifically using second morning voids, we may have underestimated the performance of the Kawasaki equation.Third, by having the three diets succeed each other without any interval, there could be a carry-over effect.However, by using the urine samples collected on day seven of each week for the main analysis, we expect this carryover effect to be minimal.Fourth, although multiple non-consecutive 24-h urine samples are considered the reference standard for measuring sodium and potassium excretion, in this study we used a single 24-h urine, this might have underestimated the agreement between estimated and measured 24-h urinary sodium and potassium excretion.Lastly, to assess whether the 24-h urine collections were complete, we did not use the recommended reference standard of monitoring urinary excretion of p-aminobenzoic acid (PABA) administered in tablet form at intervals during the urine collection [44].Instead, we assessed total urine volume and the difference between the expected 24-h creatinine excretion and the actual measured values of 24-h urinary creatinine excretion which is considered inferior to the PABA method [44].
In conclusion, three commonly used equations that estimate 24-h urinary sodium and potassium excretion (as a proxy for dietary intake) with the use of spot urine sodium and potassium measurements showed substantial bias, poor precision and poor accuracy.Use of these formulas to monitor (changes in) salt intake is therefore not recommended.The Na/K ratio based on multiple casual urine samples may be a useful, low-burden, low-cost alternative method to 24-h urine collection for monitoring daily salt intake.

Figure 1 .
Figure 1.Change in the spot urinary sodium levels over time during three different sodium diets as determined using a monitoring device.Time-series boxplot of spot sodium levels (in mmol/L) during intermediate sodium diet (3.3-5.0 g sodium/day), low sodium diet (<3.3 g sodium/day) and high sodium diet (>5.0 g sodium/day).

Figure 2 .
Figure 2. Change in the spot Na/K ratio over time during three different sodium diets as determined using a monitoring device.Time-series boxplot of spot Na/K ratio during intermediate sodium diet (3.3-5.0 g sodium/day), low sodium diet (<3.3 g sodium/day) and high sodium diet (>5.0 g sodium/day).

Figure 3 .
Figure 3. Correlation between measured and estimated 24-h sodium excretion.Scatter plots with regression lines and Kendall rank correlation coefficients for measured versus estimated 24-h urinary sodium excretion by the Kawasaki method (A), Tanaka method (B), INTERSALT method (C) and for the spot Na/K ratio and 24-h Na/K ratio (D).

Figure 4 .
Figure 4. Agreement between measured and estimated 24-h urinary sodium excretion.Bland-Altman plots presenting measured versus estimated 24-h urinary sodium excretion using the Kawasaki method (A), the Tanaka method (B), the INTERSALT method (B) and spot Na/K ratio versus 24-h Na/K ratio (D).The dashed line in the middle is the mean difference.The upper and lower limits of agreement (dashed lines) are the mean difference ± 1.96 Â standard deviation.

Figure 5 .
Figure 5. Absolute (A) and relative (B) difference distributions of measured and estimated 24-h urinary sodium excretions.

Table 1 .
Schedule of urine collections.

Table 2 .
Characteristics of the study population.

Table 4 .
Misclassification of the three estimation methods for individual sodium excretion.