Contribution of Water from Food and Fluids to Total Water Intake: Analysis of a French and UK Population Surveys

Little has been published on the contribution of food moisture (FM) to total water intake (TWI); therefore, the European Food Safety Authority assumed FM to contribute 20%–30% to TWI. The aim of the present analysis was to estimate and compare TWI, the percentage of water from FM and from fluids in population samples of France and UK. Data from 2 national nutrition surveys (Enquête Comportements et Consommations Alimentaires en France (CCAF) 2013 and the National Diet and Nutrition Survey (NDNS) 2008/2009–2011/2012) were analyzed for TWI and the contribution of water from FM and fluids. Children and adults TWI were significantly lower in France than in the UK. The contribution of water from foods was lower in the UK than in France (27% vs. 36%). As TWI increased, the proportion of water from fluids increased, suggesting that low drinkers did not compensate by increasing intake of water-rich foods. In addition, 80%–90% of the variance in TWI was explained by differences in water intake from fluids. More data on the contribution of FM to TWI is needed to develop more robust dietary recommendations on TWI and guidance on fluid intake for the general public.


Introduction
With increasing recognition of the relationship and understanding of the mechanism between water and beverage intake, hydration and health [1][2][3][4], it is important to develop evidence-based recommendations and guidelines for water intake. Recommendations, such as those published by the European Food Safety Authority (EFSA) on adequate intakes (AIs) of total water intake (TWI) [5], form the basis of public health strategies, intervention programs and food based dietary guidelines. Recommendations on the AI of total water are based partly (e.g., EFSA) [5], or wholly on population survey intake data (e.g., Institute of Medicine (IOM)) [6]. It is important to emphasize that these recommendations are for TWI, i.e., water from fluids (drinking water and water from all other beverages) and water from foods (food moisture (FM)).
The water content of foods is highly variable. For example, cucumber and lettuce are approximately 96% water, whole, boiled chicken eggs are 75% water, while digestive (semi-sweet) July 2013. To control for seasonal differences in intake, the study was carried out in four successive phases (October-December, January-March, April-mid-June, mid-June-July), during each of which approximately a quarter of the participants were included.
A seven-day diary was used to assess all dietary intake (fluids and solids). The participants received written instructions on how to complete the seven-day food and fluid diary. The diary was used to record information on all eating and drinking occasions throughout the day and for individual eating occasions. The circumstances of intake (time and location) were noted. Respondents completed the food diary with the aid of a validated photographic booklet [22], which presented various common foods and beverages in different portion sizes. Investigators ensured completeness of the seven-day diary through face-to-face interviews. The water content of reported intakes was obtained from a recognized French food composition table [23].  [24] and all seasons were sampled.
One adult (aged 19 years and over) and one child (aged 1.5-18 years) were randomly selected from individual households. A sample was taken from a list of all addresses in the UK and clustered into Primary Sampling Units (PSUs) based on small geographical areas, which were randomly selected from postcode sectors. A sample of 21,573 addresses was then randomly selected from the PSUs, which had also been randomly selected. A total sample of 4156 individuals aged 1.5 and over (including 386 young children aged 1.5-2 years, 1687 children aged 3-18 years and 2083 adults aged 19 years and over) completed three or four dietary recording days for years 1-4 combined. To ensure comparability between age groups of both surveys, data from children aged less than four years was excluded in the NDNS sample. Some individuals later agreed to be interviewed by a nurse with fewer agreeing to give blood or urine samples. The data were weighted to minimize selection bias. All relevant research ethics and governance committees approved the survey.
Subjects were asked to complete a four-day estimated food diary with a randomly selected start date, which facilitated sampling across all days of the week. An interviewer visited the household to explain how to keep the four-day food diary. Interviewers visited or phoned the household on day two or three to check the ongoing diary completion and visited to review and collect the diary no more than three days after the final recording day. Photograph atlases were provided to aid estimation of portion sizes. Parents or carers assisted children aged 11 years and younger with completion of the diary. A trained nurse measured height and weight on a subsequent visit.
Completed diaries were coded and entered into HNR's dietary assessment system (DINO, Diet In Nutrients Out) that used food composition data from the Department of Health's NDNS Nutrient Database to estimate dietary intake of foods, including water and beverages and nutrients, including energy and water. Computerized raw data files and documentation from this survey were obtained under license from the UK Data Archive [25].
The food groups considered as fluids are shown in supplementary materials Table S1. Soup was categorized as a food and milk as fluid.

Statistical Analysis
The SAS 9.2 software (SAS Institute Inc., Cary, NY, USA) was used for statistical analysis. The data of TWI were normally distributed. Continuous variables are presented as mean and standard deviation as appropriate and dichotomous variables as frequency and percentage. The number and proportion of participants with an inadequate TWI was calculated by comparing TWI to the age and gender specific AI of TWI set by EFSA [5], which are shown in Table 1. General linear models were used to identify the variation in TWI due to water from fluids or FM. Before analyzing the four waves of NDNS as one sample, a statistical comparison between the waves was performed. A significant difference in TWI between waves was observed only among subjects aged 11-18 years (p = 0.006), and not among subjects aged 4-10 years, 19-64 years and ≥65 years. Since the difference in TWI between consecutive waves was on average 63 mL/day and not consistent across age groups, aggregating the four waves for analysis was considered appropriate.
Students' t-test was used for comparative statistics; a p-value < 0.05 was considered significant. These data were normally distributed and are reported as mean (SD). Data of water intake were not normally distributed and are reported as median (interquartile ranges (IQR)).

Results
There were significant differences in demographic data for the two countries as shown in Table 2. French children were significantly lighter (p = 0.002 and 0.001 for males and females, respectively); however, French children had significantly higher energy intakes (p = 0.0007 and < 0.0001 for males and females respectively). Similarly, UK adults were significantly heavier than French adults despite reporting significantly lower energy intakes.

Contribution of Water from Food Moisture and Fluids to Total Water Intake According to Country, Gender and Age Groups
TWI was significantly higher in males compared with females in France (p = 0.0055) and the UK (p < 0.0001) as shown in Table 3. Mean TWI in French children was significantly lower than UK children (p = 0.019). The contribution of water from fluids was consistently higher in the UK than in France, overall (73% vs. 64%) and for each age and gender category. In France, males had a higher proportion of water coming from fluids; however, in the UK, there was little difference between the genders for all age groups. Water intake from fluids and food moisture were significantly higher in UK adults than in French adults (both p < 0.0001). Table 3. Contribution of water from food and fluids to total water intake (TWI) according to country, gender and age groups.  Figure 1 shows TWI and the proportion of water from food and fluids split into deciles for male adults. In both countries, as TWI increased, the proportion of water from fluids increased accordingly. The finding was consistent in the female adults (supplementary materials Figure S1) and children (supplementary materials Figure S2).

Contribution of Water from Food Moisture and Fluids to TWI According to Adherence to EFSA Adequate Intake
In both countries, most children (88%-90%) had TWI below the EFSA recommendations for AI (Table 4). Fewer adults achieved the EFSA recommendations in France than in the UK; women were consistently more likely to have adequate intakes than men. In France, those who achieved the recommendations drank more than those who did not: children drank about 629 mL more and  Figure 2 shows the relationship between TWI and water from food or fluids in adults (male and female) in France (a) and the UK (b). There was a higher correlation between TWI and water from fluids than between TWI and water from FM for adults in France (R 2 = 0.8002 vs. R 2 = 0.3098) and the UK (R 2 = 0.9088 vs. R 2 = 0.2203). These regressions suggested that 80%-90% of the variance in TWI was explained by differences in water intake from fluids. In children, the variance in TWI was explained by differences in water from fluids and to a lesser extent with water from foods (R 2 = 0.7612 vs. R 2 = 0.4766 in France and R 2 = 0.8868 vs. R 2 = 0.2699 in the UK; supplementary materials Figure S3).

Contribution of Water from Food Moisture and Fluids to TWI According to Adherence to EFSA Adequate Intake
In both countries, most children (88%-90%) had TWI below the EFSA recommendations for AI (Table 4). Fewer adults achieved the EFSA recommendations in France than in the UK; women were

Contribution of Water from Food Moisture and Fluids to TWI According to Adherence to EFSA Adequate Intake
In both countries, most children (88%-90%) had TWI below the EFSA recommendations for AI (Table 4). Fewer adults achieved the EFSA recommendations in France than in the UK; women were consistently more likely to have adequate intakes than men. In France, those who achieved the recommendations drank more than those who did not: children drank about 629 mL more and adults 804-884 mL more, representing a 6% to 7% higher contribution of water from fluids than from food. In the UK, there was a mean difference of 920-986 mL in water from fluids between those who achieved the recommendations and those who did not. This represented a 10% higher contribution of water from fluids to TWI than from food. Table 4. Contribution of water from food moisture and water from fluids to total water intake of children (3-18 years), male and female adults (≥19 years) in France and UK according to intakes above or below the EFSA adequate intake (AI) of total water [5].

Discussion
Given the increasing interest in water intake and health, it is important that population studies be designed and executed with due consideration of the methodological issues inherent in recording fluids. In addition to reporting, detailed information on TWI and its sources is essential. Therefore, this study aimed to analyze and report the contribution of water from fluids and food moisture to TWI in two nationally representative surveys. Independently of gender, adults (≥19 years) in the French survey had lower TWI intakes compared with UK adults. This finding is mirrored in studies of total fluid intake (TFI) with French adults having lower intakes compared with those in the UK [26]. The TWI observed in the 2008/2009-2011/2012 NDNS is comparable to the analysis of the 2002/2001 NDNS data, which showed that adult men had a TWI of 2.53 ± 0.86 L/day and women a TWI of 2.03 ± 0.71 [27]. The TWI of children was comparable to analyses of other TFI data sets that also showed lower TFI in French children compared with the UK [28]. The TWI for French children was comparable to data from another French national survey (INCA) conducted in 2006/2007 [29]. Data on TWI in children in the UK from other analyses have not been published to date.
Regardless of the difference in TWI between both surveys, many children and adults in both countries had TWI below the EFSA (2010) AIs. This was especially true for children, nearly 90% of whom had a TWI lower than the AI. However, the median intakes for UK adults were close to the recommendations. The French results are similar to those of Vieux et al. who reported that 89%-93% in children aged 4-13 years had intakes less than EFSA's AIs [29]. No comparable studies are available on TWI for the UK, although Iglesia et al. [28] and Gandy [30] have reported that 30%-56% of children and adolescents did not achieve the AIs when estimating TFI alone. Despite these differences, the present study and previous analyses show that the observed median intake of the sample is lower than the age-and gender-specific AI. However, when making such comparisons, it is important to consider the purpose of the AIs and how they were developed, and to only draw generalizable conclusions. The AIs are partly based on population intakes and are designed for population comparisons not for use in individuals. It is also important to note that, in the absence of hydration biomarkers, an intake below AI does not automatically equate to underhydration or dehydration. However, there is increasing evidence that, in children, dehydration and an increased fluid intake may impact cognitive and physical performance, and overall well-being [31][32][33]. Therefore, it would seem prudent to include biomarkers in future surveys in order to establish if those with reported low TWIs are at risk of mental and/or health issues.
There are undoubtedly many reasons for the differences in TWI between French and UK adults including climate, physical activity, assessment method and psychosocial factors. It is doubtful that these differences could be explained by climate, as while Northern France has a climate similar to the UK, the south is generally hotter than the UK. Likewise, since activity levels are low in both samples, they are unlikely to be explained by differences in physical activity levels. In addition, differences in methodologies may have contributed to these differences, even though both surveys used food diaries. Social factors are likely to explain some, if not most, of the variation. Dietary intake patterns vary between the countries. For example, Bellisle et al. estimated that, in France, 80%-87% of TFI was drunk during meals [34], and similarly Vieux et al. estimated that nearly 70% of all fluids are being consumed during three principle meals [29]. In the UK, the opposite was observed in a fluid specific survey; up to 70% of beverages were consumed outside of meals [30], when there are obviously more opportunities to consume fluids.
The importance of fluid intake to obtain a higher TWI was confirmed in the present analysis when looking at the contribution of FM to TWI. Indeed, the regression analyses demonstrated that differences in TWI were predominately due to differences in the amount of water from fluids. This was confirmed by the analysis of the ratio of FM and water from fluids by deciles of TWI; as the contribution of water from fluids to TWI increased 13%-20% between the first and tenth deciles in male adults in France and UK, respectively. However, the volume of this increase appeared more important (e.g., difference in male adults 1532 mL in France and 2991 mL in the UK). This analysis also suggests that participants with a low water intake from fluids did not compensate by having a higher intake of FM.
The mean percentage contribution of water from FM to TWI in the UK sample was consistently lower due to more fluids being consumed than among the French sample-27% and 36%, respectively. Similar values have been reported from earlier surveys in France: using data from the 2002/2003 CCAF survey Bellisle et al. estimated FM to be 36%-38% in French children and 36%-41% in French adults [34]. Similarly, Vieux et al. estimated FM to be 40% of TWI in French children aged 4-13 years [29]. In the UK, no specific data on FM have been reported for children; however, values of 23%-32% for all age groups (1.5-64 years) [35] and 25% in adults [27] have been reported, values in line with those in the present analysis. In other European countries, FM levels of 33%-38% have been reported in German children [36] and 32% for 9-75 year olds in Spain [37]. O'Connor has reported a FM of 33% for adults in the Republic of Ireland [38]. Therefore, the current analysis and the existing literature seem to confirm the EFSA's assumption that FM contributes for 20%-30% to TWI. However, this ratio may not be applicable to all European countries [5]. It would be beneficial if future national food surveys record and report both TWI and the contribution of water from FM and fluids. More primary data published on this topic will enable the development of robust dietary recommendations on TWI. Moreover, further exploration of the contribution of specific foods, especially high water content foods such as fruit and vegetables, to total FM may yield additional insight into these country differences, as would an exploration of social differences between countries regarding drinking and eating behavior.
The present analysis has several strengths. Both surveys used an estimated food diary, a method that is considered to be the most robust available dietary assessment method [39]. Using surveys with the same assessment method was a selection criterion since different methods could create a bias in the results. For example, fluids can be underestimated when intake is assess with a 24 h recall and, consequently, this could bias the ratio of water from FM/water from fluids [40]. The collection of data across all seasons in both surveys will have minimized seasonal variations in food choices e.g., variation in type and quality of fruit and vegetables or fluids consumed. Such variations would have influenced the findings in this study due to variations in FM.
However, some limitations in the present analysis must be acknowledged besides those inherent in dietary surveys. While both surveys are considered representative of the respective country's populations, the sampling methods were slightly different. The data from the UK is a summary of four years of a rolling program as opposed to France's one year survey, which is repeated on a three-year basis. This resulted in significantly different sample sizes, with over twice as many subjects being surveyed in the UK compared with France. Additionally, in France, a seven-day diary was used while a four-day diary was used in the UK. However the four-day diary used by NDNS was validated against the previously used seven-day diary, and, therefore, is considered as representative and comparable [41]. Any dietary survey is subject to potential errors inherent in the methodology, for example underreporting. In this analysis, energy intake was lower than in the UK than in France, despite UK children and adults being heavier. Interestingly, TWI and its sources were lower in France, clearly demonstrating the need for methodologies validated for specific nutrients.

Conclusions
The findings, together with the current literature, indicated that TWI and the contribution of water from FM and fluids were variable between countries. Also within one country, TWI as well as the contribution of water from FM varied greatly. More publications of primary data analysis from more countries on this topic are needed to develop more robust dietary recommendations on TWI and FBDG, preferably on fluid intake in addition to TWI. Water from fluids was shown to be the main driver of TWI. Encouraging the consumption of fluids, especially drinks that do not contribute to total energy intake such as water, seems therefor appropriate to increase TWI to the levels recommended by EFSA.  Table S1: the categorization and codes of fluids used in analysis of the CCAF survey and NDNS survey.