Comparison of three methods for estimating daily individual ...

European Journal of Clinical Nutrition (1999) 53, 281±287 ß 1999 Stockton Press. All rights reserved 0954±3007/99 $12.00 http://www.stockton-press.co.uk/ejcn

Comparison of three methods for estimating daily individual discretionary salt intake: 24 hour recall, duplicate-portion method, and urinary lithium-labelled household salt excretion A Melse-Boonstra1, H Rexwinkel1, J Bulux2, NW Solomons2 and CE West1* 1

Division of Human Nutrition and Epidemiology, Wageningen Agricultural University, Wageningen, The Netherlands; and 2Center for Studies of Sensory Impairment, Aging and Metabolism (CeSSIAM), Guatemala-City, Guatemala

Objective: To compare methods for estimating discretionary salt intake, that is, salt added during food preparation and consumption in the home. Setting: The study was carried out in a rural Guatemalan village. Subjects: Subjects were selected non-randomly, based on their willingness to cooperate. Nine mother-son dyads participated; the sons were aged 6 ± 9 y. Interventions: Three approaches for estimating the discretionary salt consumption were used: 24 h recall; collection of duplicate portions of salt; and urinary excretion of lithium during consumption of lithium-labelled household salt. Total salt intake was assessed from the excretion of chloride over 24 h. Results: The mean discretionary salt consumption based on lithium excretion for mothers was 3.9 2.0 g=d (mean s.d.) and for children 1.3 0.6 g=d. Estimates from the 24 h recalls and from the duplicate portion method were approximately twice and three times those measured with the lithium-marker technique respectively. The salt intake estimated from the recall method was associated with the lithium-marker technique for both mothers and children (Spearman correlation coef®cient, 0.76 and 0.70 respectively). The mean daily coef®cient of variation in consumption of discretionary salt measured by the three methods, for mothers and boys respectively, were: lithium marker, 51.7 and 43.7%; 24 h recall, 65.8 and 50.7%; and duplicate portion, 51.0 and 62.6%. Conclusions: We conclude that an interview method for estimating discretionary salt intake may be a reasonable approach for determining the relative rank-order in a population, especially among female food preparers themselves, but may grossly overestimate the actual intake of salt added during food preparation and consumption. Descriptors: salt consumption; methods; lithium-marker technique; 24 h recall; duplicate portion method; Guatemala

Introduction Sodium in foods can originate from various sources, some is naturally present in foods, some is added during the industrial preparation and processing of foods, often in the form of sodium chloride, while some is discretionary being added during food preparation and consumption in the home (cooking and table salt). In studies of the relationship between salt intake and hypertensive vascular disease (Intersalt Cooperative Research Group, 1988; Khaw & Barrett-Connor, 1988; Tuomilehto et al, 1980; Alderman et al, 1998) or in diet therapy for patients with renal insuf®ciency, hepatic diseases or congestive heart failure (Thomas, 1994), emphasis on the total dietary delivery of sodium is the focus of attention. The discretionary component of sodium chloride *Correspondence: Dr CE West, Division of Human Nutrition and Epidemiology, Wageningen Agricultural University, PO Box 8129, 6700 EV Wageningen, The Netherlands. Received May 25 1998; revised August 25 1998; accepted November 17 1998

is of special interest. Not only is it a contributor to total sodium intake but, given the almost universal and constant consumption of salt, it is also a potential vehicle for delivering micronutrients. It has extensively been used for delivery to populations of iodine (Venkatesh Mannar, 1994), and has been considered for other nutrients, such as ¯uoride (Wespi, 1989) and iron (Sayers et al, 1974), as well. To assess the intake of salt, various methods have been used in the past. In individuals with normal renal function, the 24 h excretion of sodium in urine provides an accurate approximation of the intake the previous day (Bates & Thurnham, 1997). To determine the proportion of dietary sodium which is discretionary, traditional methods of dietary assessment including 24 h recalls or records (Pietinen, 1982; Holbrook et al, 1984; Caggiula et al, 1985; Mattes & Donnelly, 1991; Olubodun et al, 1997), dietary history (Cerqueira et al, 1979; Van Dokkum et al, 1989) or food frequency questionnaires (Sowers & Stumbo, 1986; Pietinen et al, 1982; O'Donnell et al, 1991) have been used. Although the person preparing the food knows how much salt is added during food preparation, others in the

Estimation of discretionary salt intake A Melse-Boonstra et al

282

household consuming the food are usually unaware of the amount of salt added. Duplicate portion techniques are well established for assessing total nutrient intake and are considered to be one of the most accurate methods available for assessing habitual dietary intake (Bingham et al, 1995). The method has been used to assess the intake of sodium and a number of trace elements (Clark & Mossholder, 1986; Anderson et al, 1993). Because of the dif®culty of recalling the discretionary use of salt, especially in cooking, we adapted the duplicate portion method for estimating household salt use. To our knowledge, this is the ®rst time this method has been applied in the ®eld to questions of sodium chloride consumption in the home. Another method is that of `household salt disappearance'. With this method, the amount of salt or foods containing salt entering the household as gift, purchase and=or barter over a speci®c time period is estimated by questionnaire. Using a factor to estimate intra-householdmember distribution, an estimate of individual intake can be calculated (Pietinen, 1981; Edwards et al, 1989). Based on salt purchase, the average salt intake in Guatemala has been estimated to be 11 and 9 g per capita per day in rural areas and the capital, respectively (Ministerio de SaluÂd Publica y Asistencia Social, 1996). Recognizing that all dietary assessment methods have their limitations in both precision and accuracy, the validity of two of the methods mentioned above Ðthe 24 h recall and the duplicate portion methods Ðwere of interest to us in the context of ®eld testing the method for tracking discretionary salt intake with lithium-labelled salt via its excretion in quantitative collections of urine (SanchezCastillo et al, 1984, 1987a; James et al, 1987; Leclerq et al, 1990). While exploring this method in a rural, Guatemalan village as part of a multicenter collaboration between Guatemala and Benin (Melse-Boonstra et al, 1998), we have compared the values registered for total discretionary salt intake using 24 h recall, duplicate portion, and lithium labelled salt. The values provided by the various approaches have been correlated and cross-calibrated. Subjects and methods Location The study was conducted in Buena Vista, a village of about 1700 inhabitants in the municipality of San Pedro SacatepeÂquez at 2500 m above sea-level in the central highlands

Figure 1 Experimental design.

of Guatemala. The village is 2 km from the township San Pedro, which is 20 km from the center of the Republic's capital, Guatemala City. The climate is temperate and dry, with a mean temperature of 16 C. Ninety percent of the population is of indigenous origin; most women wear traditional dress while men wear western clothes. The indigenous inhabitants speak Cakchiquel, one of the 20 Mayan dialects spoken in the country. Some people, especially women, do not speak Spanish. Most women are illiterate. Selling of agricultural products and handmade clothes are the main sources of income. There is one bakery, and about 15 small shops (tiendas) which sell basic groceries. Preparation of the salt The lithium-labelled salt given to the subjects was prepared by AKZO-Nobel (The Netherlands), according to the method described by Sanchez-Castillo et al, 1987b. Grain size was set on 300 ±800 mm, as the general salt in Guatemala is coarse. The spread in lithium content was checked, after packing the salt in 40 plastic bags of 250 g, by taking samples of 10 g from every fourth bag and analysing it at Wageningen Agricultural University. The samples showed a mean content of 1.575 g Li=kg of salt (CV: 3.7%). The 10 kg lithium labelled salt available for this study was calculated to be suf®cient to study salt consumption in nine families. Subjects Nine Buena Vista mothers and their sons, aged 6 ±9 y, participated in this study. They were selected primarily on their willingness to participate, therefore, a highly cooperative study population was enrolled. Consent of the mothers for themselves and their sons was requested and provided only after they had been given a full explanation of the study in their own language. The Human Studies Committee of CeSSIAM approved the study protocol. Experimental design The experimental design is shown schematically in Figure 1. The experimental period lasted 10 d for each family. Each week on Wednesday, three families commenced with the study. In order to limit disturbance of family life, no major activities were carried out during the weekend. For all families, urine and salt collections took place on Tuesday, Wednesday and Thursday, while the 24 h recalls were carried out on Wednesday, Thursday and Friday in order to


assess food intake on the collection days. The ®eldwork took place in January and February, 1996. For the lithium marker technique, on day 0, that is, one day before the start of the intervention the subjects collected urine for 24 h in order to provide a baseline population-average for the excretion of lithium, sodium and chloride. The ®rst urine voided on rising was not collected, but all subsequent urine voided over the next 24 h period, including the ®rst urine voided on the following day, was collected. Collection of baseline urine was important for two reasons. Firstly, since water and some vegetables may be a source of lithium (Sanchez-Castillo et al, 1987c), we were able to correct for any lithium that was not provided by the intervention. However, no correction for baseline lithium excretion was made because it was negligible ( < 0.005 mmol=d). Secondly, the baseline excretion of sodium and chloride would enable any major variation to be detected in overall behaviour related to dietary sodium sources produced by the complex intervention of salt replacement, urine collections and frequent interviews. During the experimental days, the nine families used only the lithium-labelled salt instead of their usual household salt. The investigators collected all regular salt in the households so that no unlabelled salt was available in the household during the study. At the conclusion of this study the salt was returned to the families. On days 7, 8 and 9, 24 h urine samples were collected. Urine collections were planned and timed so as to avoid the dates of the next expected menstrual period of the mother. No internal or external marker for the completeness of urine collection was employed because we regarded the extra burden for the participants as not acceptable. Instead, we chose to rely on careful questioning of the mother regarding any breaks in protocol. When any doubt as to the quantitative certainty of a full 24 h collection arose, such samples were not analysed. A total of 16 urine-days (22%) in 11 subjects were deleted based on the incompleteness-of-collection criterion. As all children went to school during the day, the help of the headmaster and teachers of the village school was sought. Teachers were informed which children were scheduled to collect urine. Small personal collection bottles were left with the teachers and given to the children when they left the classroom to go to the toilet. Concentrations of lithium and sodium were determined with an atomic absorption spectrophotometer (Perkin-Elmer 2380, Norwalk, Connecticut, USA), and urinary chloride was measured using a coulometric method with a Chloor-o-counter (type 77, Marius, Utrecht, The Netherlands). All measurements were carried out in duplicate. For the interview method, 24 h recalls were carried out with salt as a speci®c topic in the interview. Subjects were ®rst interviewed at baseline and then one day after each urine collection day, that is, on days 8, 9 and 10. We asked the female head of household about food preparation for the whole family and about the individual consumption of foods and beverages by the mother and by her index child on the previous day. Salt in food or beverages consumed outside the home was not regarded as discretionary, therefore no attempt was made to estimate this intake. The method was a combination of an interview with weighing and measuring the amounts of foods which were indicated as having been eaten (MenchuÂ, 1991). A trained Guatemalan nutritionist-dietician carried out the interviews, accompanied by a Cakchiquel-speaking translator from the Buena Vista community.

For the duplicate portion method, the person in charge of the cooking was instructed to place the same amount of salt used during the cooking in a labelled plastic box (family salt box) each time food was prepared. In our study the mothers in the households were solely responsible for cooking. In addition, mother and child were asked to put the same amount of salt that they added to their food during the meal into a labelled plastic box (individual salt box) brought to the dining area. In order to limit changes in behaviour with respect to the addition of salt, subjects were instructed to maintain their usual behaviour, using either their hands or utensils for adding salt to the food and placing it in the duplicate portion boxes. The contents of each box were collected and weighed on a digital scale to the nearest gram. The proportion of the salt in the family salt box consumed by mother and her child was estimated from the 24 h recall data. Mathematical calculations Assuming that dietary chloride not originating as sodium chloride was negligible, total salt (NaCl) intake was calculated from the urinary output of chloride. Daily discretionary salt consumption was calculated for each individual by dividing urinary lithium excretion by the proportion of lithium in the salt (1575 mg Li=kg salt; CV: 3.7%). Individual means were calculated from data of those days that urine collection was complete. From these data, group means for mothers and of children were calculated. Mean intake of energy, fat, protein, carbohydrate and sodium was calculated using food composition tables for Mexico (MunÄoz de ChaÂvez et al, 1996) and Latin America (INCAP-ICNND, 1961). The Mexican tables were used because sodium content of foods is not included in the Latin American food composition databases. The food composition table of Mexico used for this study contained analyses of the sodium content of some foods but was incomplete for a number of the foods reported in Buena Vista. This produces an intrinsic underestimation to the value provided for total sodium intake relying on food composition tables. Statistical analyses All statistic calculations were carried out using SAS, version 6.09 (SAS Institute Inc., 1989). All variables were checked for normality using tests for skewness, Kurtosis, Shapiro=Wilk, Stem Leaf and Normal probability plots. Descriptive statistics were calculated for each variable. As the data were normally distributed (with a number of minor exceptions), means of the estimates obtained with the lithium marker technique (reference method) were compared with those obtained by 24 h recall and double portion method using the paired Student's t-test. Spearman correlation coef®cients were used to relate the reference method with the other two methods. Results The mean age of the mothers was 37.4 y (median: 40 y; range: 25 ±46 y) and of the children 8.1 y (median: 8 y; range: 6 ±9 y). Mean volume of 24 h urine collections was 1.10 0.24 L (mean s.d.) (median: 1.12 L; range 0.58± 1.70 L) for mothers and 0.56 0.24 L for children (median: 0.61 L; range 0.21 ±1.27 L). The difference in mean urine volume between mothers and children was statistically signi®cant (P < 0.001). Of the 72 urine collections, 56

283


284

Table 1 Comparison of estimates of discretionary salt intake as measured by three methods Discretionary salt (g=d) Mothers (n 9) Mean s.d. Lithium-marker technique 24 h recall Cooking salt Table salt Total salt Double portion method Cooking salt Table salt Total salt

Children (n 9) CV (%)

Mean s.d.

CV (%)

**

43.7

3.9 2.6

51.7

6.2 4.4 1.2 1.9 7.4 4.9

71.7 158.8 65.8

3.3 1.6 0.4 0.8 3.7 1.9*{

48.4 201.0 50.7

6.9 4.0 5.9 6.1 12.8 6.5{

57.6 103.5 51.0

3.9 1.9 3.3 4.5 7.1 4.5{

48.2 137.3 62.6

Comparison by Student t-test with values for mothers: * P < 0.05;

**

1.3 0.6

P < 0.01, and with values obtained by lithium marker technique: { P < 0.05.

Table 2 Daily intake of macronutrients and sodium (from other sources than household salt) determined by 24 h recall of mothers and children compared with the RDA at moderate physical activity Mothers (n 9)

Energy (MJ) Protein (g) Fats (en%) Carbohydrates (en%) Sodium (mmol) a b

Children (n 9)

Day 0

Mean day 7 ± 9

Recommended

8.8 3.0 70.5 31.6 11.5 4.9 77.8 25.8 39.5 70.4b

8.0 1.8 68.1 15.9 11.6 3.8 76.6 18.8 21.6 15.3

8.8 53 20 ± 25 60 ± 70 21.8

a

Day 0

Mean day 7 ± 9

Recommended a

5.5 1.9 43.0 14.5 13.6 7.7 75.4 27.8 22.7 39.1b

5.4 1.0 45.2 12.8 14.6 4.9 73.1 13.5 14.4 11.6

7.9 32 20 ± 25 60 ± 70 17.4

From: MunÄoz de ChaÂvez M et al (1996). High baseline value is due to a large value in one mother-son pair.

(78%) were complete according to self-report criteria. The results described are based on data from the complete collections. The average discretionary salt consumption, as determined by the lithium-marker technique (Table 1), was found in the mothers to be 3.9 2.0 g=d (mean s.d.) (median: 3.4; range: 0.7 ±9.4 g=d) while for children, 1.3 0.6 g=d (median: 1.3; range: 0.4 ±3.1 g=d); mean values for mothers and children being signi®cantly different (P < 0.01). According to the 24 h recall method, the discretionary salt consumption was 7.4 4.9 g=d (median: 6.3; range: 2.2± 17.2 g=d) for mothers and 3.7 1.9 g=d (median: 3.4; range: 2.2 ±8.2 g=d) for children, a signi®cant difference between groups (P < 0.05). Ingestion of energy, protein, carbohydrate, fat and sodium from non-discretionary sources (intrinsic to foods) for mothers and children are listed in Table 2. The duplicate-portion method showed higher mean values for discretionary salt intake, namely 12.8 6.5 g=d (median: 10.1; range: 8.5 ± 28.0 g=d) for mothers and 7.1 4.5 g=d for children (median 6.2; range: 2.4 ± 16.9 g=d). The maternal-offspring difference disappeared with this approach to calculating intake (P > 0.05). Minor breaches in the protocol were detected; these included that: (1) for one mother the portion meal size could not be calculated with the 24 h recall; (2) two families did not collect duplicate portions on the ®rst collection day; and (3) one family did not collect duplicate portions at all. Values for discretionary salt consumption obtained by 24 h recall were 90% greater, on average, than the estimate given by the lithium-marker method for the mothers, and 185% greater for the boys. Use of table salt according to

24 h recall is 16% and 11% of total discretionary salt consumption for mothers and children, respectively, with salt added in cooking constituting the remainder. Compared to the 24 h recall reporting on salt use, the duplicate portion method gave values that were approximately 73% and 92% higher for mothers and children, respectively. The duplicate-portion values, moreover, were 228% greater, on average, than the estimate given by the lithium-marker method for the mothers, and 446% greater for the boys, Table 1 provides a summary of these ®ndings. According to the duplicate portion method 46% of the discretionary salt is used as table salt. Due to the broad inter-individual variability within a small population of nine family units, the mean discretionary salt consumption of mothers using the 24 h recall method did not differ signi®cantly from the results of the lithium-marker technique, although the outcome of the 24 h recall method is almost two-fold higher; for children, the inter-method difference was signi®cant (P < 0.02). Estimates from the duplicate-portion method were signi®cantly different for both the maternal and the sons' groups as compared to the lithium-marker technique, respectively (P < 0.03 and P < 0.02). In terms of rank-order associations, for mothers the Spearman correlation coef®cient for the 24 h recall vs the lithium-marker technique was 0.76 (P < 0.01). For the duplicate-portion method vs the lithium-marker technique it was also 0.76 (P < 0.01) in the maternal group. For children, the coef®cients were 0.70 (P < 0.05) and 0.54 (P < 0.05), respectively. The inter-person variation of discretionary salt consumption is expressed in the coef®cients of variation (relative standard deviation) of the mean salt consumption


Table 3 Day-to-day variation in salt consumption for mothers and children as measured by three methods of estimation Individual CV, mean % (range)

Lithium-marker technique 24 h recall method Double portion method a

Mothers

Children

36.8 (14.6 ± 52.7; n 8)a 47.9 (17.5 ± 108.6; n 9) 34.7 (10.0 ± 72.9; n 9)

44.0 (2.6 ± 82.5; n 6)a 43.5 (10.3 ± 108.6; n 9) 54.2 (14.2 ± 82.8; n 9)

If urine collections were missing for one day the CV was calculated over two days.

(Table 2). For mothers, these were 51.7%, 65.8% and 51.0% according to the lithium-marker technique, 24 h recall method and duplicate portion method, respectively, and for children 43.7%, 50.7% and 62.6%. Day-to-day variation in salt consumption was 36.8% for mothers and 44.0% for children, according to the lithium-marker technique (Table 3). Except for household salt few other sources of salt in the daily food intake were found in this community. Most important of these is a seasoning for beef or chicken, packed in little sachets. Other sources are cookies (sweetened) and crackers (salted), crisps and foods bought outside the home, such as tamales or chuchitos (foods made from maize dough baked in broad leaves and usually stuffed with meat or chicken). Discussion Many methods are used to measure the intake of foods and nutrients by individuals and populations but, where possible, methods should be calibrated against or validated in terms of a biologically-based `gold standard' method (Dyer et al, 1997; Kaaks, 1997; OckeÂ & Kaaks, 1997; Bingham & Day, 1997). We believe that the lithium-label method has been established as a suitable, biological reference method for tracking the actual consumption of household salt by individuals (Leclerq et al, 1990). Our ®ndings indicate that the 24 h recall and duplicate portion method provide a rankorder for individuals within a population that is relatively consistent with the actual relative hierarchy as established by a gold-standard method. Hence for correlation analysis or for relative-risk studies, these techniques are useful. For the estimation of the total amount of salt from food prepared and consumed in the home these methods may provide a gross overestimation. The estimates obtained from the 24 h recall were twice those obtained with the lithium label method. From this it could be inferred that more than 100% of all excreted sodium would have come from discretionary sources. With the duplicate portion method, estimates of discretionary salt intake were three times those provided by the reference method. Such values were higher than the total amount of sodium excreted in urine. As pointed out by Sanchez-Castillo et al (1984), imperceptible losses of the salt between the source of the salt supply and the target food occur in both the kitchen and at the table. Assuming that the duplicate amount of salt added to the salt boxes was representative of that being added to cooking pots and to meals in Buena Vista, nearly two-thirds of this salt never passes the lips. This accounts for part of the over-estimation of discretionary salt intake by the dietary methods. This is possibly combined with a tendency to overestimate, rather than underestimate, the visual

perceptions of salt being used in terms of that actually being added in a discretionary manner. Therefore, while taking advantage of the acceptable ®delity of the ranking of a population's relative rank-order of discretionary salt consumption, one must be extremely cautious with the use of the absolute values for estimated total intake of salt from discretionary sources as provided from interview methods. In terms of customary intake of salt from discretionary sources, people vary greatly one from another. The coef®cients of variation of > 50% for all methods attest to this. Day-to-day variation in household salt consumption was in the 37% (mothers) to 44% (children) range, for three consecutive days, Tuesday to Thursday. This is comparable to the 36% for variation observed in the Italian population by Leclerq et al (1990). By not using an objective criterion for completeness of urine collection we took the risk of underestimating the real salt consumption of our subjects. Mean urine volume and ion excretion for mothers was consistent over the 4 d of urine collections; for children the urine volume and ion excretion had decreased signi®cantly during the experimental days compared to the baseline (Melse-Boonstra et al, 1998). However, even when assuming an under collection of 40%, a discretionary salt intake of 1.8 g=d is still only half of the salt consumption as measured with the 24 h recall and a quarter as measured with the double portion method. At the group level, no signi®cant difference in salt consumption between days was shown, according to the lithium marker technique. As the measurements were always made on the same days (Tuesday through Thursday) and the missing data were evenly distributed over days we think the breaches in the protocol did not lead to systematic error in the comparisons. Our primary programmatic interest in Guatemala with intake of discretionary salt had less to do with chronic diseases or clinical dietetics than with the problem of iodine de®ciency disorders (IDD). IDD is endemic to Guatemala, and despite legislation mandating the iodization of table salt, a substantial prevalence of IDD remains (MartõÂnez, 1988). Implementation of the legislation calls for commercial salt to contain from 30 ± 100 mg=kg (ppm) of iodine. The national median in a national survey lay between 15 ± 30 mg=kg (ppm) (Ministerio de SaluÂd Publica y Asistencia Social, 1996). With the incomplete iodization shown by this national survey, and con®rmed by another Guatemalan study (Stewart et al, 1996), it would require a daily intake of 4 ± 8 g of household salt by children and 5± 10 g by adults to achieve the recommended intake of iodine from this source, which is 120 mg and 150 mg iodine per day for children and adults, respectively (Melse-Boonstra et al, 1998).

285


286

As planners and policy makers might base their judgements on the iodine protection of the population by indirect assessments of salt intake based on interviews, it was essential to test the validity of these approaches. The various methods in current use have a number of pitfalls and caveats. The household disappearance method often assumes a uniform distribution within the family unit. As much research with alternative methods (MartõÂnez-Salgado et al, 1990, 1993; Engle & Nieves, 1993) consistently shows a lower food intake by children than by adults consistent with their differential physiological needs, the assumption of uniform intra-family partition is erroneous. More importantly, the household disappearance method assumes that household members consume all salt entering the home. The method does not take into account salts consumed by pets or used for other purposes. Interview methods do not take into account losses of salt during food preparation. Sanchez-Castillo & James (1995) found that only 15± 35% of cooking salt is retained when vegetables are cooked in salted water. Secondly, estimation of salt intake from food consumption data requires reliable data on the sodium content of foods. In the third place, in a largely illiterate population, interview methods may be less applicable than in literate populations. Conclusions We have examined the ability of the 24 h recall and duplicate portion method to determine the use of discretionary salt in Guatemalan homes in relation to a biologically-based `gold standard' method. Within the constraints of interpretation from the small sample population size that we could enrol, we conclude that 24 h recall and duplicate portion methods overestimate the actual intake of salt from discretionary sources. The results of this small study might provide a reasonable approach for examining associations between high and low consumption of this condiment with health. To examine such associations, a wider study is however required. Acknowledgements ÐWe want to thank Dr Ivania Mena and the mothers' committee of Buena Vista for all the help they gave during the ®eldwork. We also want to thank Ana Luisa GuilloÂn from NEO (New Educational Opportunities) for pretesting the 24 h recall forms. A special thanks also to Lucy Santiago for carrying out the 24 h recalls. From the Wageningen Agricultural University we want to thank Ir. Paul Hulshof and Frans Schouten for analyzing the urine samples. For making available lithiumlabelled salt we thank Ir. Monique Veenendaal from AKZO-Nobel, Hengelo, The Netherlands.

References Alderman MH, Cohen H & Madhaven S (1998): Dietary sodium intake and mortality: the National Health and Nutrition Examination Survey (NHANES I). Lancet 351, 781 ± 785. Anderson RA, Bryden NA & Polansky MM (1993): Dietary intake of calcium, chromium, copper, iron, magnesium, manganese and zinc: Duplicate plate values corrected using derived nutrient intake. J. Am. Diet. Ass. 93, 462 ± 464. Bates CJ & Thurnham DI (1997): Biomarkers of nutrient intake. In: Design Concepts in Nutritional Epidemiology, 2nd Edn, B Margetts & M Nelson, (eds). Oxford: Oxford University Press, pp 170 ± 240. Bingham SA, Cassidy A, Cole TJ, Welch A, Runswick SA, Black AE, et al. (1995): Validation of weighed records and other methods of dietary assessment using the 24 h urine nitrogen technique and other biological markers. Br. J. Nutr. 73, 531 ± 550. Bingham SA & Day N (1997): Using biochemical markers to assess the validity of prospective dietary assessment methods and the effect of energy adjustment. Am. J. Clin. Nutr. 65, (Suppl), S1130 ± S1137.

Caggiula AW, Wing RR, Nowalk MP, Carole Milas N, Lee S & Langford H (1985): The measurement of sodium and potassium intake. Am. J. Clin. Nutr. 42, 391 ± 398. Cerqueira MT, McMurry Fry M & Connor WE (1979): The food and nutrient intakes of the Tarahumara Indians of Mexico. Am. J. Clin. Nutr. 32, 905 ± 915. Clark AJ & Mossholder S (1986): Sodium and potassium intake measurements: dietary methodology problems. Am. J. Clin. Nutr. 43, 470 ± 476. Dyer A, Elliott P, Chee D & Stamler J (1997): Urinary biochemical markers of dietary intake in the INTERSALT Study. Am. J. Clin. Nutr. 65 (Suppl), S1246 ± S1253. Edwards DG, Kaye AE & Druce E (1989): Sources and intakes of sodium in the United Kingdom diet. Eur. J. Clin. Nutr. 43, 855 ± 861. Engle PL & Nieves I (1993): Intra-household food distribution among Guatemalan families in a supplementary feeding program: behavior patterns. Soc. Sci. Med. 36, 1605 ± 1612. Holbrook JT, Patterson KY, Bodner JE, Douglas LW, Veillon C, Kelsay JL, et al. (1984): Sodium and potassium intake and balance in adults consuming self-selected diets. Am. J. Clin. Nutr. 40, 786 ± 793. Institute of Nutrition of Central America and Panama Ð Interdepartemental Committee on Nutrition in National Defence (INCAP ± ICNND) (1961): Tabla de compocision de alimentos para uso en America Latina. INCAP, Guatemala City, Guatemala. Intersalt Cooperative Research Group (1988): Intersalt: an international study of electrolyte excretion and blood pressure. Results for 24 hour urinary sodium and potassium excretion. Br. Med. J. 297, 319 ± 329. James WPT, Ralph A & Sanchez-Castillo CP (1987): The dominance of salt in manufactured food in the sodium intake of af¯uent societies. Lancet 1, 426 ± 429. Kaaks RJ (1997): Biochemical markers as additional measurements in studies of the accuracy of dietary questionnaire measurements: conceptual issues. Am. J. Clin. Nutr. 65 (Suppl), S1232 ± S1239. Khaw KT & Barrett-Connor E (1988): The association between blood pressure, age, and dietary sodium and potassium: a population study. Circulation 77, 53 ± 61. Leclercq C, Avalle V, Ranaldi L, Toti E & Ferro-Luzzi A (1990): Simplifying the lithium-marker technique to assess the dietary intake of discretionary sodium in population studies. Clin. Sci. 79, 227 ± 231. MartõÂnez MM (1988): De®ciencia de yodo, bocio endemico y su relacion con el estado nutricional en escolares de la Republica de Guatemala. MS theseis in Food, Nutrition and Health, University of San Carlos, Guatemala, Center of Superior Studies in Nutrition and Food Science, Guatemala. MartõÂnez-Salgado H, Molina Rivera E, ChaÂvez Villasona A & Mata Estrada A (1990): Variaciones en el consumo individual de sal en presencia o ausencia de chile en la comida, en una comunidad de MeÂxico. Arch. Latinoam. Nutr. 40, 45 ± 54. MartõÂnez-Salgado H, Tovar-Zamora E, ChaÂvez-Villasana A, ArmendaÂriz DM & Baz-DõÂaz-Lombardo G (1993): Consumo familiar e individual de sal de mesa en el Estado de MeÂxico. Sal. PuÂbl. MeÂx. 35, 630 ± 636. Mattes RD & Donnelly D (1991): Relative contributions of dietary sodium sources. J. Am. Coll. Nutr. 10, 383 ± 393. Melse-Boonstra A, Rozendaal M, Rexwinkel H, Gerichhausen MJW, Briel vd T, Bulux J, et al. (1998): Determination of the consumption of household salt in order to determine the level of iodine forti®cation of salt required to control iodine de®ciency disorders in rural Guatemala and Benin: Studies using lithium-labelled salt. Am. J. Clin. Nutr 68, 636 ± 641. MenchuÂ MT (1991): GuõÂa metodologica para realizar encuestas familiares de consumo de alimentos. Institute of Nutrition of Central America and Panama, Guatemala City, Guatemala. Ministerio de SaluÂd Publica y Asistencia Social (1996): Encuesta nacional de micronutrientes. Institute of Nutrition of Central America and Panama, Guatemala City, Guatemala. MunÄoz de ChaÂvez M et al (1996): Tablas de valor nutritivo de los alimentos de mayor consumo en MeÂxico. EdicioÂn internacional, Editorial Pax MeÂxico, MeÂxico. OckeÂ MC & Kaaks RJ (1997): Biochemical markers as additional measurements in dietary validity studies: application of the method of triads with examples from the European Prospective Investigation into Cancer and Nutrition. Am. J. Clin. Nutr. 65 (Suppl), S1240 ± S1245. O'Donnell MG, Nelson M, Wise PH & Walker DM (1991): A computerized diet questionnaire for use in diet health education. 1. Development and validation. Br. J. Nutr. 66, 3 ± 15. Olubodun JOB, Akingbade OA & Abiola OO (1997): Salt intake and blood pressure in Nigerian hypertensive patients. Int. J. Cardiol. 59, 185 ± 188. Pietinen P (1981): Sources of sodium in the Finnish diet. J. Sci. Agric. Soc. Finland 53, 275 ± 284.


Pietinen P (1982): Estimating sodium intake from food consumption data. Ann. Nutr. Metab. 26, 90 ± 99. Pietinen P, Tanskanen A & Tuomilehto J (1982): Assessment of sodium intake by a short dietary questionnaire. Scand. J. Soc. Med. 3, 105 ± 112. Sanchez-Castillo CP, Seidell J & James WPT (1987a): The potential use of lithium as a marker for the assessment of the sources of dietary salt: cooking studies and physiological experiments in men. Clin. Sci. 72, 81 ± 86. Sanchez-Castillo CP, Branch WJ & James WPT (1987b): A test of validity of the lithium-marker technique for monitoring dietary sources of salt in man. Clin. Sci. 72, 87 ± 94. Sanchez-Castillo CP, Warrender S, Whitehead T & James WPT (1987c): An assessment of the sources of dietary salt in a British population. Clin. Sci. 72, 95 ± 102. Sanchez-Castillo CP, Warrender S, Whitehead T & James WPT (1984): A novel technique for assessing the sources of dietary sodium. Ann. Clin. Res. 16 (Suppl 43), S44 ± S48. Sanchez-Castillo CP & James WPT (1995): De®ning cooking salt intakes for patient counselling and policy making. Arch. Latinoam. Nutr. 45, 259 ± 264. SAS Institute Inc. (1989): SAS=STAT User's Guide, Version 6, 4th Edn, Volume 1 and 2, Cary, NC: SAS Institute Inc.

Sayers MH, Lynch SR, Charlton RW, Bothwell TH, Walker RB & Mayet P (1974): The forti®cation of common salt with ascorbic acid and iron. Br. J. Haematol. 28, 483 ± 495. Sowers M & Stumbo Ph (1986): A method to assess sodium intake in populations. J. Am. Diet. Ass. 86, 1196 ± 1202. Stewart C, Solomons NW & Mendoza I (1996): Salt iodine variation within an extended Guatemalan community. The failure of intuitive assumptions. Food Nutr. Bull. 17, 258 ± 261. Thomas B (ed) (1994): Manual of Dietetic Practice. Edited for the British Dietetic Association. Cambridge, UK: University Press. Tuomilehto J, Karppanen H & Vuori J (1980): Sodium and potassium excretion in a sample of normotensive and hypertensive persons in eastern Finland. J. Epidemial Community Health 34, 174 ± 178. Van Dokkum W, De Vos RH, Muys Th. & Wesstra JA (1989): Minerals and trace elements in total diets in The Netherlands. Br. J. Nutr. 61, 7 ± 15. Venkatesh Mannar MG (1994): The iodization of salt for the elimination of iodine de®ciency disorders. In SOS for a Billion, eds. BS Hetzel & CS Pandev. Oxford: Oxford University Press, 89 ± 107. Wespi HJ (1989): Development, present status, and recommendations for improvement in caries prevention with ¯uoridated salt in Switzerland. Schweiz. Monatsschr. Zahnheilkd. 79, 851 ± 859.

287