Introduction

Stroke used to be the leading cause of death between 1951 and 1980, and is still one of the three major causes of death in Japan.1,2 Several epidemiological studies have clarified that Japan's high stroke incidence is mainly due to a high prevalence of hypertension.2,3 Therefore, the main focus of anti-cardiovascular disease programmes has been preventing stroke by lowering blood pressure, and the core of these programmes has not changed substantially. Even now, hypertension is one of the most important risk factors for cardiovascular diseases in Japan.2

There are several known factors that increase blood pressure, such as excessive salt intake, obesity, heavy drinking, lack of exercise and mental stress.4,5,6,7 On the other hand, it has been reported that sufficient intake of potassium (K) may decrease blood pressure.8

The Japanese salt intake used to be markedly higher than that in western countries in the 1980s.5 According to Japan's National Nutrition Survey 1998, the mean salt intake has been decreasing every year.9 However, salt intake in Japan is still higher than in western countries,5 and it is still important to further decrease salt intake and increase K intake. The national health guideline from 2001 to 2010, ‘Health Japan 21’, has set goals for salt and potassium intake, and programmes to achieve them are conducted by local municipalities.10 In order to evaluate salt and potassium intake in community-based programmes or in epidemiological surveys, we need a simple method of estimating population salt and potassium intake.

There are several methods for evaluating salt and potassium intake, such as dietary recall or records from 24 to 96 hours, food frequency questionnaires and 24-h urine collection. Generally, 24-h urine collection is considered to be the most reliable method to evaluate these intakes.11,12,13,14,15 However, this method involves a considerable burden on subjects and it is difficult to collect complete and accurate 24-h urine samples. This method may not be practical in public health practice or in large-scale epidemiological surveys. Kawasaki et al12,13 reported a method of estimating 24-h urinary sodium (Na) and K excretion from second morning voiding urine (SMU) in adults. This method is fairly accurate and is used in some epidemiological studies. However, in many local municipalities or working places, urine collection in medical check-ups is not restricted to the morning, which makes it difficult to take SMU from all subjects.

The purpose of our study is to develop a simple method of estimating the 24-h urinary Na (24HUNaV) and K (24HUKV) excretion from the Na/Cr or K/Cr ratio (where Cr stands for creatinine) of a spot urine specimen collected at any time. The method will be very useful to estimate the population level of Na (or salt) and K intake for epidemiological surveys and public health activities aimed at reducing blood pressure levels in a community, and to compare their annual trends.

Materials and methods

We developed a formula to estimate 24-h urinary Na and K excretion, using Japanese data items from the INTERSALT study, which was conducted through a highly standardised method, as a gold standard.5 In this study, participants were asked to provide both a casual ‘spot’ urine specimen and a 24-h urine collection sample. We developed formulas by comparing both data sets.

The design of the INTERSALT study has been detailed elsewhere.14,15,16 Briefly, 52 centres from 32 countries participated in the study. In Japan, three populations from Osaka, Toyama and Tochigi were studied between 1987 and 1988. A total of 200 subjects including men and women aged 20–59 years were randomly selected by each centre. We obtained complete data for 200, 197 and 194 subjects from Toyama, Osaka and Tochigi, respectively. Subjects were first asked to urinate and empty their bladder completely and provide a ‘spot’ urine specimen. Then, they were asked to start 24-h urine collection. The ‘spot’ urine specimen was not included in the 24-h urine collection, but its electrolyte and Cr concentrations were analysed as a casual urine specimen. All specimens were refrigerated at 4°C within 24 h and frozen at −20°C within 7 days, then shipped in frozen state to the central laboratory in St Rafäel University, Louvain, Belgium. Na and K were examined by emission flame photometry and Cr by the Jaffé method.

Although there is no satisfactory biological method for determining the completeness of a 24-h collection, subjects were given verbal and printed explanations on the need for collecting all urine over 24 hours, along with instructions for avoiding several common errors that lead to omission of a small part of the specimen. At the end of the collection, completeness was assessed through a non-judgmental interview.

Since it is known that 24-h urinary Cr excretion can be affected by age, height and weight, we used forward stepwise regression analysis to calculate a regression equation to estimate 24-h urinary Cr excretion (24HUCrV) from age, weight and height.17,18,19,20

We developed formulas to estimate 24-h urinary Na (24HUNaV) and K (24HUKV) using a similar method reported by Kawasaki et al.12,13 The estimation procedure is shown in Table 1: equation (1): 24-h urinary Cr excretion (24HUCrV) can be predicted, and the predicted value of the Cr (PRCr) is almost equal to the measured 24HUCrV (Kawasaki et al12,13), equation (2): The ratio of Na or K to Cr in the 24-h urine is directly proportional to the ratio of Na or K to Cr concentration in the ‘spot’ urine specimen (SU). We assumed that if equations (1) and (2) in Table 1 can be proved, then the proportional expression (3) will be derived from them. Equation (3): 24-h urinary Na or K excretion (24HUNaV or 24HUKV) is directly proportional to the value of the ratio of Na or K to Cr in SU multiplied by PRCr.

Table 1 Hypothesis to estimate 24-h urinary Na and K excretion
Full size table

To examine the accuracy of these equations, we applied them to an external group. A total of 513 manual workers aged 20–69 years who did not participate in the INTERSALT study were asked to provide both a ‘spot’ urine specimen and a 24-h urine collection. Complete data were obtained from 280 men and 56 women. We compared the estimated values to the measured values by the same method.

Then, the external group was divided into quintile groups according to the estimated 24HUNaV or 24HUKV. We compared estimated values of 24HUNaV and 24HUKV from these equations to measured values from 24-h urine collection in quintile categories.

The Statistical Package for the Social Sciences (SPSS. version 10.0J; SPSS Japan, Tokyo, Japan) was used for the analysis. Linear forward stepwise regression analysis was employed to calculate regression equations to predict 24HUCrV. Comparisons between estimated values and measured values by 24-h urine collection were carried out by Student's or Welch's t-test. All P values were two-tailed.

Results

The baseline characteristics of the subjects are shown in Table 2.

Table 2 Number of subjects, age, body height, body weight, body mass index, urinary volume, 24-h urinary excretion of sodium, potassium and creatinine by sex, in Japanese aged 20–59 years in the INTERSALT study, 1987–1988
Full size table

The proportions of subjects providing a ‘spot’ urine specimen in the morning, and in the afternoon and evening, were similar (51.8% vs 48.2%).

Using linear forward stepwise regression analysis, we established a regression equation to predict 24HUCrV from age, weight and height. The obtained formula predicting PRCr was as follows; PRCr = −2.04 × age + 14.89 × weight (kg) + 16.14 × height (cm) − 2244.45.

The correlation between 24HUNaV (or 24HUKV)/ 24HUCrV ratio and SUNa (or SUK)/SUCr ratio is shown in Figure 1. The correlation coefficient (R) was 0.65 for Na and 0.67 for K (P < 0.01).

Figure 1
figure 1

Relationship between (a) the 24HUNaV/24HUCrV ratio and the SUNa/SUCr ratio, and (b) the 24HUKV/24HUCrV ratio and the SUK/SUCr ratio in 591 Japanese subjects in the INTERSALT study aged 20–59 years in 1987–1988. 24HUNaV or 24HUKV = 24-h urinary Na or K excretion (mEq/day); SUNa or SUK = Na or K concentration (mEq/L) in the spot urine; SUCr = creatinine concentration (mg/dl) in the spot urine.

Full size image

Since we assumed hypotheses (1) and (2) were proved, we estimated 24HUNaV and 24HUKV according to the hypothesis (3).

In Figure 2, the XNa or XK terms calculated from the formula (XNa (or XK) = SUNa (or SUK)/SUCr × PRCr) were significantly correlated with the measured 24HUNaV (or 24HUKV). The regression lines of the estimated 24HUNaV (YNa) or the estimated 24HUKV (YK) were; YNa = 0.29 × XNa + 121.0 (r = 0.53), YK = 0.27 × XK + 28.9 (r = 0.54). If the concentration in SUNa or SUK nearly reaches zero, 24HUNaV or 24HUKV will theoretically reach almost zero, and the predicted value from the regression line will have to coincide with the measured value. Therefore, these values were transformed into common logarithms similar to a previous report (Kawasaki et al21) and from them we created formulas to estimate 24-h Na or K excretion. These formulas are expressed as the estimated 24HUNaV (Y′Na) = 21.98 × XNa0.392 and the estimated 24HUKV (Y′K) = 7.59 × XK0.431. Figure 3 shows the associations between the estimated values from Y′Na or Y′K and the measured values of 24HUNaV and 24HUKV. They were significantly associated with each other, and the correlation coefficient (R) was 0.54 for Na (P < 0.01) and 0.56 for K (P < 0.01). We compared the estimated values of 24HUNaV or 24HUKV from these formulas to measured values in the subjects of the INTERSALT study. The mean ± standard deviation (s.d.) of the estimated 24HUNaV was 178.9 ± 36.2 mmol/day, and the mean ± s.d. of the measured 24HUNaV was 187.2 ± 65.8 mmol/day. The mean ± s.d. of estimated 24HUKV was 44.0 ± 9.0 mmol/day, and the mean ± s.d. of measured 24HUKV was 45.8 ± 16.0 mmol/day. There was a significant difference between the estimated and the measured values in Na and K excretion. However, the difference between the measured and the estimated values was fairly small; 8.3 mmol/day in 24HUNaV or 0.5 g/day in salt, and was 1.8 mmol/day in 24HUKV.

Figure 2
figure 2

Relationship between the value from equation(3) (XNa or XK) and (a) the 24UNaV or (b) 24HUK measured in 591 Japanese subjects in the INTERSALT study aged 20–59 years in 1987–1988. Formula XNa (or XK) = SUNa(or K)/SUCr × PRC r; 24HUNaV or 24HUKV = 24-h urinary Na or K excretion (mEq/day).

Full size image
Figure 3
figure 3

Relationship between (a) 24-h urinary Na or (b) K excretion estimated and measured in 591 Japanese subjects in the INTERSALT study aged 20–59 years in 1987–1988. Estimated 24HUNaV = 21.98 × ((SUNa/(SUCr × 10)) × PRCr)0.392; Estimated 24HUKV = 7.59 × ((SUK/(SUCr × 10)) × PRCr)0.431; SUNa or SUK = Na or K concentration (mEq/L) in the spot urine; SUCr = creatinine concentration (mg/dl) in the spot urine.

Full size image

We created separate formulas by the collected urine time for the morning and afternoon. The difference between the estimated value and the measured value in 24HUNaV was 7.2 mmol/day in the morning and 8.2 mmol/day in the afternoon. The difference in 24HUKV was 1.7 mmol/day in the morning and 1.8 mmol/day in the afternoon.

In the external population, we compared the estimated values of 24HUNaV or 24HUKV from these formulas to the measured values in Table 3. There was a significant difference between the estimated and measured values in Na and K excretion. However, the difference between the measured and the estimated values was 24.0 mmol/day in 24HUNaV, in terms of sodium chloride; 1.4 g/day, and 3.8 mmol/day in 24HUKV.

Table 3 Mean and correlation coefficients of measured and estimated 24-h urinary Na, K, Creatinine in the external population, 280 men and 56 women, aged 20–69 years (the mean ± standard deviation)
Full size table

We divided these external group subjects into five groups according to each quintile category by the estimated 24HUNaV or 24HUKV, and Table 4 shows the estimated and measured values in each quintile. The difference between the estimated and measured values in 24HUNaV was from 0.5 to 45.4 mmol/day, from 0.03 to 2.7 g/day for sodium chloride, and from 0.1 to 10.7 mmol/day for 24HUKV. The estimated 24HUNaV and 24HUKV were parallel to the measured 24HUNaV or 24HUKV.

Table 4 Mean and standard deviation of estimated and measured 24HUNaV and 24HUKV in each category of quintile by the estimated 24HUNaV or 24HUKV
Full size table

Discussion

In this study, we clarified the validity of a simple method to estimate the population mean of Na and K excretion using casual (‘spot’) urine specimens.

There are established methods to estimate Na and K excretion, each of which has advantages and disadvantages. The 24-h urine collection method is considered to be the most reliable, because most of the Na and K a subject takes is excreted into the urine, as long as there is no external loss such as diarrhoea or excessive perspiration.22 However, the 24-h urine collection method is not readily conducted in ordinal population groups. Many subjects are not willing to collect their urine for a whole day in their daily living, because they always have to carry urine jars with them and they need to be careful to collect their urine completely. There is a report that the rate of unsuccessful collection is about 40%.23 In order to make urine collection easier, Tochikubo et al24 developed a portable, semiautomatic urine sampling device with divided partitions. This device is small enough to be carried in a plastic bag and can collect 1/50 amount of the total urine excreted during 24 hours. However, even with this kind of device, it is likely that subjects will forget to collect their specimens. Another method is to collect a part of each subject's specimens and estimate the population average of Na and K intake from them. Hirata et al25 and Kawasaki et al12,13 developed methods for estimating 24-h urinary Na and K excretion from spot urine or division urine. The method of estimating them from a second morning voiding urine specimen developed by Kawasaki et al12,13 is actually used in some epidemiological studies and health education. However, collecting SMU from all subjects is usually very difficult, because medical check-ups are not always carried out in the morning. If the medical check-ups are conducted from morning until evening, the urine collection time also varies.

In order to resolve the disadvantages of the available methods, we developed a simpler way of estimating population means of Na and K excretion utilizing a casual (‘spot’) urine specimen that is collected in medical check-ups with no restriction on collection time. The collecting time of the spot urine we used ranged from 8.00 am to 7.00 pm. In the pre-existing methods, the spot urine specimens have to be collected from second morning voiding urine, urine at night or urine accumulated for a specific period of time. The 24-h urinary Na and K excretion are affected by circadian rhythms; they are low at night and become high from the afternoon to the evening.26 Urinary Na and K excretion at night are largely influenced by the amount of daytime exercise27 and salt intake.28 In our estimation procedures, we developed equations from Na/Cr ratios of 24-h urine and spot urine. According to Kawasaki et al,29,30 in this method, the correlation between the estimated value and measured value is high when either SMU or urine accumulated from 8.00 am to 12.00 am is used, and that the correlation is low when using night urine.29,30 Thus, the correlation between the estimated and the measured value depends on the timing of urine collection, and it is apparent that we should use urine collected within limited times to estimate individual Na or K excretion. Although we created separate formulas by the collected urine time for the morning and afternoon, the difference between the measured and estimated value was very similar, and if the formulas for estimating vary according to the timing, the estimation of population 24-h urinary Na and K excretions would become complex and impractical. Even if the urine collection timing is not limited and personal accuracy decreases, the correlation coefficient between the estimated and measured values is fairly high for Na (r = 0.54) and for K (r = 0.56). Therefore, our method would be a convenient and accurate way to estimate population Na and K intake.

Regarding urine Cr excretion estimation, Kawasaki et al19,20 developed separate formulas for each of men and women, because the amount of muscle was considered to be different by gender. However, we developed a single formula for both men and women, after clarifying separate formulas and finding fair consistency in the single formula estimation of urine Cr excretion.

As for examining the validity of the estimated formulas, we applied them to the external group. Statistically significant correlation coefficients were obtained for both Na and K. However, there were significant differences between the mean values of estimated and measured values in Na and K. Because 24-h urinary Na excretions have a circadian rhythm and the collecting time was distributed randomly for all day, some cases may have shown the large differences between the estimated and the measured values. But in all subjects, the difference was small.

In an additional validity examination, when the subjects were divided into quintile groups by the estimated value, the means of the measured values were higher in the groups with higher estimated values. This indicates that we can estimate the change of the actual Na or K excretion by obtaining the change in the estimated values from the spot urine. However, the difference between measured and estimated values was larger in quintile groups with a smaller amount of Na or K excretion. This may be because the regression model assumed a curvature passing the origin, which resulted in a larger discrepancy as the estimated value approached zero. This method is nonetheless considered to be useful to estimate Na excretion of subjects with relatively high salt intake, because the differences were smaller in the quintile groups with larger amounts of Na excretion.

In summary, we established a simple method for estimating the population mean of Na and K excretion from casual (‘spot’) urine specimens and examined the validity of the estimation. The method is considered to be available for comparing different groups, populations and annual trends of a particular group in health education and on other occasions. However, since this method is not suitable for individual subjects, for estimating the personal Na and K intake or alternation of Na and K intake before and after the intervention or education, not this method but other methods, for example reported by Kawasaki et al12,13 or 24-h urine collection, should be used. ‘Health Japan 21’ has set a goal for salt intake for the whole Japanese population to less than 10 grams per day and various kinds of intervention programmes have been conducted by each municipal government. Monitoring and evaluation are required to achieve their goals, and our method will be useful in estimating the population mean of the salt intake.