Prediction of 24-Hour Urinary Sodium Excretion Using a Single Spot Urine Samples in Moroccan Population

Background: Excessive sodium intake is linked to high blood pressure. Estimating sodium intake is difficult. The 24-h urine collection is currently the recommended method for estimating intake but cumbersome for large population studies. Predictive model to estimate sodium intake based on single spot urine were developed, but showed inconsistency when used in extern populations. This study aims to develop a specific model for estimating sodium excretion over 24 hours for the Moroccan population. Methods: 371 participants in the urinary validation sub-study of the STEP-wise survey-Morocco 2017-2018 provided a valid 24-hour urine collection and spot urine specimens. Participant were randomly assigned to the training (n=183) and the validation data set (n=188). Results: A prediction model for 24-hour sodium excretion was developed. Adjusted R2 was 0.258. In the validation data set, correlation was 0.431 [95%CI; 0.258-0.580], and the adjusted R2 was 0.190. The Bland-Altman plot showed a nonsignificant small mean bias of -18 mg (95%CI, -213 to 177) in predicting 24-h urinary sodium excretion at the group level. At the individual level, limits of agreement were wide. Conclusion: This new model developed from a single spot urine could be used to predict the average 24-h sodium excretion of Moroccan adults.

Thus, there is an urgent need to find more alternative methods to 24-hour urine collection as the WHO deadline approaches [12]. Equations to predict 24-hour urinary sodium excretion (24HUNa) [13][14][15][16] have been developed based on spot urinary sodium and creatinine concentrations as well as basic characteristics of individuals such as age, height and weight and/or body mass index (BMI). Recently, numerous studies have examined the validity of previously published predictive equations for estimating 24HUNa across populations. Investigators found poor agreement between predicted and measured 24HUNa, suggesting the inadequacy of these predictive equations for use in a population different from that in which they were initially developed [17][18][19][20][21]. Thus, it is recommended to develop new population-specific equations. We aim here to develop and validate a Moroccan specific predictive equation for 24HUNa .

Study design and population
Data were based on the urinary validation sub-study of the STEP-wise survey-Morocco 2017-2018. Briefly, the STEP-wise survey-Morocco 2017-2018 was a nationwide cross-sectional study that aimed to assess the prevalence of the risk factors of NCDs in Moroccan adults. A total of 6,100 adults were selected to participate in the STEP-wise survey-Morocco 2017-2018 using the standard WHO STEPS sampling design which is a robust probability-sampling process that allows for the selection of representative data with three sampling stages (cluster, household, and individual). Data were collected in three steeps: 1) clinical interview including self-reported information on basic characteristics and health behavioral risk factors of participants, 2) blood pressure (BP) and anthropometric measurements, and 3) biochemical analysis (blood and urine). Details on the STEP-wise survey-Morocco 2017-2018 were described elsewhere [22]. A total of 371 individuals (139 Men and 253 women) aged 18 years and over who reside in the rural and urban areas of Rabat-Salé-Kenitra region (the North West region of Morocco) were included in the urinary validation substudy using the same sampling design of the STEP-wise survey-Morocco 2017-2018[22]. Participants were asked to provide a 24-hour urine collection and a spot urine sample. Individuals declared to have kidney diseases or engaged in a salt-free diet, and women in menstruation period were instructed not to collect urine and were excluded from this study. The field work was conducted during March-June 2017. This study was approved by the Biomedical Research Ethics Committee at the Faculty of Medicine and Pharmacy of Rabat. All participants have signed informed consent.

Urine collection
Detailed oral and written instructions for 24-hour urine collections and spot urine specimens were presented to individuals who were eligible and willing to participate in the study. The 24-h urine samples were collected according to the WHO/PAHO standard protocol [23]. Participants were asked to discard the first void collected in the morning and keep collecting all urine voided during the rest of the 24-hour period, including the first void of the second day. The time of the start and the finish of the 24-hour collection was recorded by each participant. Participants were instructed to store the urine samples in a dark and cold place. Once finished, the 24-h urine collections were transported in a thermoelectric coolers to the laboratory.
The total volume of the 24-h urine collection was recorded. After shaking, 5-ml aliquots were prepared and immediately frozen at -21° for analysis at the laboratories of "Joint Research Unit in Nutrition and Food" Rabat. The evening spot urine specimens were collected one day before the 24-h urine collection as a fraction of a void collected during the evening period between 7 p.m. and 10 p.m. The spot urine specimens were treated using the same methods for the 24-hour urine samples. The completeness of the 24-hour urine collection in the present study was defined as: (1) total urine volume 400 ml ≤ V ≤ 3600 ml; (2) collection period 22h ≤ time ≤ 26h (3) participant reported losing more than few urine drops or missing one void.
This approach was used in previous references largescale international epidemiological studies such as INTERALT, INTERMAP and NHANES. Sodium and potassium concentrations were analyzed using the Inductively Coupled Plasma Mass Spectrometry (ICP-MS; Thermo Scientific XSERIES2). An international reference material, Seronorm TM Trace Elements Urine, was used to control and validate the measurements. The creatinine excretion was analyzed using the Jaffe method using the Cobas C311 (Roche diagnostic, Meylan-France). The sodium and creatinine concentrations in the 24-h urine collection were adjusted by the total volume of the collection. The coefficients of variation for sodium, potassium and, creatinine were all less than 3%, indicating a low analytical imprecision.

Weight, Height and Blood Pressure
Physical measurements included weight, height, and BP were recorded by a health professional according to the WHO protocol [24] (page 182-184). Height was measured to the nearest 0.5 cm and weight was measured to the nearest 0.1 kg. Body mass index (BMI) was estimated as weight in kilograms divided by height in meter squared (kg/m 2 ). Three brachial BP measurements were recorded using a calibrated digital sphygmomanometer (Digital Automatic Blood Pressure Monitor DABPM Spengler ES 60). Three brachial systolic blood pressure (SBP) and diastolic blood pressure (DBP) records were averaged to determine the mean BP values. All participants completed 3 measurements. Blood pressure was measured in participants in sitting position and appropriate cuff sizes was used. Before BP measurements, participants were rested for at least 5 min. Detailed procedures and equipment's were described in the STEPS manual published by the WHO [25].

Statistical analyses
Data was partitioned randomly (stratified by sex and age) into training (n=183) and validation (n=185) data sets.

Developing a Moroccan prediction model
We developed a regression model to predict 24-hour urinary sodium excretion using individuals from the training data set. Similarly to the INTERSALT equation [13], our model was developed using one-step regression, which is an approach expected to reduce the amount of measurement error compared to the two-steps approach used to develop Kawasaki, Mage and Tanaka equations [14][15][16]. The model was developed using the measured 24-h sodium excretion as the outcome variable. Sodium and creatinine concentrations in the spot urine, age, and BMI, were used as predictors. No interaction was found between different predictors. The inclusion of each parameters as a predictor in the model was based on their known influence on sodium excretion and their significant contribution to the model. Model fit was assessed using adjusted R 2 .

Validation of the prediction model
Regression coefficients obtained from the training data set were used to estimate the 24-h sodium excretion in the validation data set. The measured and predicted 24hour sodium excretion were presented as the mean and standard deviation (SD). Pearson correlation was calculated to assess the association between measured and predicted 24-hour urinary sodium excretion using the newly developed model, and the bias (predictedmeasured) was calculated using paired t-test. The distribution of measured against the predicted 24-hour urinary sodium excretion with the regression line (95%CI) and the line of equality (identity line) was visualized using the scatter plot. [27] was used to assess agreement between the measured and predicted 24-hour sodium excretion. The degree of mean bias was calculated with the 95% limits of agreement. The 95% limits of agreement, calculated as ±1.96×standard deviation of the mean differences should include 95% of individual bias. The Bland-Altman plot was drawn by plotting the individual bias (pred 24hUNa minus meas-24hUNa ) on the Y axis against the individual means [(pred 24hUNa plus meas 24hUNa)/2] on the X axis. A sensitivity analysis was performed to assess the effect of further potential incomplete collections using 24-hour creatinine excretion.

The Bland Altman plot [26]
Analyses were replicated after excluding participants with Meas-24HUCrea less than 3 mmol or greater than 25 mmol in women or less than 6 mmol or greater than 30 mmol in men [28]. Statistical analyses were carried out using Statistical Package for the Social Sciences (SPSS, Inc., Chicago, IL, USA) and plots were built in Xlstat 2014.

Results
Basic characteristics of the study's population are shown in table 1. Overall, mean age and BMI (SD) were 48.9 years (15.1) and 27.7 kg/m 2 (5.6), respectively. The mean 24-hour urine sodium excretion was 2847 mg/day (SD; 1372) equivalent to a salt intake of 7.2 g/day (1g of Na+ is equivalent to 2.54g NaCl) [29]. Except for mean age (p=0.042) and urinary volume (p=0.038), there are no significant differences in baseline characteristics and urinary measurements between the training data set and the validation data set (table 1). The model regression coefficients derived from the training data set are shown in Table 2.      Both genders showed a nonsignificant small mean bias between the two sodium measurements (-193 mg/day; 95%CI, -531 to 145 for men vs 8 mg/day; 95%CI, -156 to 324 mg/day for women), with wide limits of agreement for men (-2952 mg/day; 2566 mg/day) and women (-2508 mg/day ; 2676 mg/day) (data not shown). The same pattern of an overestimation trend at lower 24-h sodium excretion and underestimation at higher levels was observed in men and women (data not shown).

Fig. 2:
Bland Altman plot to assess bias produced by the newly developed model to predict 24-h urinary sodium excretion presented for the total sample.
Bias was calculated as the absolute differences (predicted minus measured) in 24-h urinary sodium excretion. The 95% limit of agreement are bias ± 1.96×SD.

Completeness of 24-hour urine collections in the present study was assessed using a standard protocol based on rigorous examination of the collection period, total urine volume and a post-collection questionnaire completed by participants. This approach has been used in the International Cooperative Study on Salt, Other Factors, and Blood Pressure (INTERSALT) [31], and recently in the National Health and Nutrition
Examination Survey (NHANES) [32]. In addition, we excluded further participants with creatinine excretion values outside the range of 6 mmol/day, 30 mmol/day for men, and 3 mmol/day, 25 mmol/day for women [28], but the results remain unchanged.

Discussion
This study was designed to develop a Moroccan-specific model to estimate 24HUNa using a single spot urine specimen. The newly developed model underestimates the mean 24HUNa at the group level by a small nonsignificant bias of -17.8 mg/day. At the individual level, differences in the predicted and measured 24HUNa was inconsistent through the level of 24HUNa, with a systematic tendency of overestimation at lower level of 24HUNa and underestimation at higher level.
Research examining the utility of spot urine as a simpler and easier to implement method to replace 24hour urine collection, which is accurately the reference method for estimating sodium intake in individual and population, are of great importance as studies suggested that the 24-hour urine collection may be a barrier to estimate sodium intake, particularly in poor countries [33].
The inclusion of spot urine sodium and creatinine concentrations, age, and BMI was based on knowledge of the influence of these variables on sodium excretion [34,35,36,37], and on their significant contribution to the model. We used a one-step (direct) regression to derive an estimate for 24HUNa, similar to the INTERSALT equation developed based on data from North American and European participants of the INTERSALT study [38].
The direct regression approach used in this study is considered more appropriate compared to the two-steps regression approach used by Tanaka, Kawasaki and Mage, which consist of estimating the sodium-tocreatinine ratio in the 24-hour urine from sodium-tocreatinine ratio in the spot urine and estimate of 24HUCrea from sex, weight, height, age, and in some cases, ethnicity, since the removal of one step should reduce statistical error in the estimation procedure [38].
Our findings show that the developed model produced a small and non-significant mean bias of underestimation, overall. The mean difference indicates an underestimation bias for men and an overestimation bias for women. The reason for the change of the bias direction between genders remain unclear and could be explained by differences in sodium intake, the way the kidney handles sodium excretion differently in men and women, or differences in the parameters used to develop this model.
We found a coefficient of correlation between the measured and predicted 24HUNa of 0.431 [95%CI, 0.258-0.580]. This was lower correlation than those reported in previous studies [39], [38]. By comparison, the Tanaka and Kawasaki equations were developed and validated in Japanese populations based on a relatively higher correlations of 0.540-0.730, but no data on agreement were reported. Assessment of the agreement in means, and the variability of differences across individuals is the ideal approach to examine the validity of a new method to be used interchangeably or to replace another method [40]. The use of correlation to assess agreement of two methods has been widely criticized [41] [42].
Our following was to examine bias using the Bland-Altman method, which is the reference method for . This was not unexpected, the sodium intake of an individual varies highly from meal to meal, and the sodium concentration in spot urine specimens has been shown to reflect the sodium content in food and beverages consumed within hours of the urine collection [48]. This diurnal intraindividual variability in sodium excretion is a source of systematic error when estimating 24HUNa using spot urine [33] [48]. This is, to our knowledge, the first study to estimate 24HUNa in Moroccan population using a model based on intrinsic coefficients. However, our study had some limitations. Firstly, completeness of 24-hour samples is the major problem of the method. We used the WHO protocol to assess sodium concentration in 24-hour urine collection [49]. Standard detailed instructions were provided and a post-collection interview was conducted with each participant to exclude eventual under or over-collected urine samples based on a doubtful period of collection, loss of urine, and urine volume. Moreover, we conducted a sensitivity analysis using a criteria for 24HUCrea (<3 mmol/day or >25 mmol/day in women and <6 mmol/day or >30 mmol/day in men [52]), and the results remain unchanged. However, it is always possible that there are still potential incomplete collections in the final sample. Ensuring the completeness of collections is crucial for an accurate estimate of sodium intake. The use of reference Biomarkers such as paraaminobenzoic acid is recommended, however, it is impractical and burdensome to be used in population studies [17], particularly in low-middle income countries. Measured creatinine or ratio of measured-topredicted creatinine rate has been used to assess completion of 24-hour urine, but there are currently no standard criteria. Although relatively constant, creatinine excretion is influenced by the differences between populations in protein intake pattern, in muscle mass, and ethnicity [17]. To our knowledge, for now there is no specific equation to predict 24HUCrea in the Moroccan population. Therefore, the use of predictive equations of creatinine developed in different sitting could wrongly distinguish collections according to their completion in our population. Secondly, in the current study, only a single spot specimen collected at the evening was used. Sodium excretion vary across the day, and studies have shown inconsistency in the validity of the most frequently used predictive equations of 24HUNa depending on the time of spot urine [50] [51]. Investigators suggested collecting spot urine at random times of the day to compensate for the potential systematic error that could be produced due to the circadian rhythm of the 24HUNa [33]. We admit that the ideal would have been to collect spot urine at different time of the day in order to choose the best model. However, the choice of evening spot urine in the present study was based on metabolic studies showed that sodium excretion is maximal during the afternoon and evening and minimal from midnight to early in the morning [52] [53] [54].

In addition to numerous validation studies [39] [55]
have shown that the highest prediction accuracy was with equations based on the afternoon and evening spot as compared to the morning or overnight spot.
Findings from the current study have an important implication for future studies to estimate sodium intake in Moroccan population and monitor the trend of its change over time, which is a key information to assess the effectiveness of strategies implemented to reduce salt intake in populations. Our study support the use of a specific newly developed model based on a single spot urine concentrations and participants' basic characteristics to approximate the mean 24-h urinary sodium excretion for population, but not for individuals.
To summarize, by using a single spot urine which is more suitable and easier to collect than 24-hour urine, the newly developed model was found reliable to approximate the mean population sodium intake. But the performance at the individual level was poor.