The results of this study showed that the SFFQ had a reasonable validity in determining the sodium or salt intake in adults, children and adolescents in comparison with the standard method of 24DRs. The validity of SFFQ was relatively good in adults as compared to the 24hUNa. In this study, SFFQ overestimated the sodium intake compared to the other two standard methods. The overestimation of nutrients by SFFQ compared to other dietary assessments has been previously reported by several studies (18, 19, 20). Reproducibility of the SFFQ for assessment of sodium intake was acceptable. Although the validity of SFFQ was good but according to the Bland-Altman plots, the current bias between SFFQ and standard methods was about 500 mg sodium which was equivalent to 1.2 gr of salt. Since the mean of salt intake per day was higher than 10 gr (21), we speculate one gram of salt intake bias is not high.
Various studies indicated a wide range of validity ranging from 0.04 to 0.91 for dietary assessment methods compared to 24hUNa (22, 23, 24). According to a pooled analysis of Freedman et al of five validation studies, the mean validity of SFFQ for estimation sodium intake was 0.16 and the underestimation of the SFFQ compared to 24hUNa was 5.6% (25). The current study found a similar cross-classification agreement and higher validity and of SFFQ for sodium assessment than Xu et al' s study among Chinese women (24), while relatively less reproducibility. This is probably because of the short interval of two-week between the two SFFQs in Xu et al' s study (24). Sasaki et al. (22) showed a less validity of SFFQ for assessment of sodium intake compared to 24hUNa than the current study (26). Consistent to our findings, Reinivuo et al. (27) reported that the classification of individuals in the same and adjacent quartiles was relatively comparable in both methods. Similarly, several studies conveyed overestimation of sodium intake by dietary assessment compared to the 24hUNa (3, 27, 28). It could be reasonable since about 90% of total sodium intake might be excreted from urine (29, 30). On the contrary, the underestimation of SFFQ observed in the Finnish study (27) might have been due to collecting urine samples on Sundays, however most recall days were weekdays. It was reflecting a more consumption of high content of sodium foods during the weekends. In the current study, we estimated discretionary salt which was added at table and cooking through questioning about the weight of salt packages, the number of households, and the period of time that each salt package is used (31, 32). In the line with previous studies, adding these questions to SFFQ improved the validity and reproducibility of the questionnaire (31, 32). Our SFFQ was also valid in estimating the contribution of major food sources in sodium intake including added salt, bread, cheese meet, fast food, canned food, nuts and seeds, sweet and soft drinks and junk food, sauce and salty vegetables due to reasonable correlation coefficients between the SFFQ and the standard method of 24DRs in both age groups.
A study among Belgian school-aged children reported that the validity of SFFQ for assessment of food intake compared to a 24DRs varied from 0.10 for potato to 0.65 for skimmed milk (33). Similar to our study, they indicated an overestimation of foods such as cereals, beverages, and dairy products (33). Fumagalli et al. similarly reported that the validity of a SFFQ assessing against 24DRs in children (5–10 yrs) ranged between 0.5–0.7 for most nutrients, however it was low and prone to overestimation of sodium (34).
Possible causes of the inconsistency can be attributed to the daily changes in 24-h urine samples, individuals’ recalling errors during dietary assessments, and lack of completeness and accuracy of food composition tables (22). Recently Titze et al. suggested that salt can be stored gradually and in a great extent in the inner layer of skin. Therefore, the homeostasis of intercellular sodium cannot be confined to the kidneys, and thus, the sodium estimates by 24-h urine collection might not be accurate (35). Further potential explanations for these dissimilarities can be due to the excretion of sodium via sweat, which varies greatly depending on the type of weather and physical activity (28). Also, underestimations in 24-urine collection can simply occur through errors in urine collection methods or the loss of urine volumes. These errors were avoided in the current and previous studies (36) by calculating 24-h creatinine/weight ratio, and questioning about the individuals’ complete urine collection. On the other hand, there are several errors related to the SFFQ such as, lack of completeness and accuracy of food composition tables, errors in individuals’ reports, different sodium content in food items, and daily alterations in diet (22). Precise measurement of sodium intake is rather challenging, due to diverse distribution of sodium in foods, widespread use of sodium compounds in processed foods as well as drinking water, and high consumption of salt at the table (32).
We used the latest food composition table with primary sources being the closest to Iranian food and cuisine. The table was enhanced by adding the sodium content of sodium-containing foods, measured in previous studies (37). Over- and underestimations in sodium intake has been similarly reported by various studies, however since the most national community-based studies examined the trend of salt intake and also have categorized the people based on nutrients such as sodium, the error in the amounts of intake, when it does not correlate with the high and low levels of intake, is negligible.
Strengths and limitations:
The first strength was encompassing two standard methods including 24DRs and 24hUNa. The second strength was the wide range of age group of study population 6 years and over. This study has also accounted for the seasonal variation, hence collecting twelve 24DRS, as a dietary standard method, during one year. The limitations of this study included recall bias of in SFFQ and 24DRs, overestimation of SFFQ and using single 24-h urine collection. Finally, a larger study population could have been a valuable asset to increase the capability and accuracy the findings.