Water is not only an essential nutrient for bodily and mental functions [1
], it is starting to be identified as one of the key elements for chronic disease prevention [4
]. In separate cohorts, lower total fluid intake [8
], lower plain water intake [4
], and lower 24 h urine volume [7
] were all associated with increased risk for chronic kidney disease, and low water intake has also been associated with new-onset hyperglycaemia [6
The adequate intake (AI) for total water intake (TWI) proposed by the European Food Safety Authority (EFSA) is 2.5 L/day for adult males and 2.0 L/day for adult females [9
]. This dietary reference intake is less restrictive than the AI established by the US Institute of Medicine (IOM) at 3.7 L/day for men and 2.7 L/day for women [10
Sources of TWI are fluid intake (sum of drinking water and all other beverages), food moisture, and metabolic water (derived from oxidation of macronutrients). Despite a general consensus on the major role of fluids [3
], the quantitative contribution of the different sources to the TWI is lacking evidence. Based on observed fluid intake data from the National Health and Nutrition Examination Survey (NHANES) III, the IOM reported that 81% of TWI came from fluids and 19% from foods [10
]. The assumption made by EFSA is that fluids contribute 70%–80% of TWI and food moisture 20%–30% [9
]. A limited number of studies in Europe reported contributions of fluids to TWI ranging from 67% in Ireland up to 75% in the UK [11
]. These ratios were means established in a population sample, and possibly masked a large variability depending on fluid intake. Documenting the contributions of water from fluids and from food moisture could be essential when translating the AI for TWI into an easy-to-understand and practical dietary guideline on fluid intake for the general population.
In France, studies describing TWI using a representative sample of the population are scarce. Drinking water was found to be the main source of fluids in all age groups [15
] with existing variations between tap water and bottled water intakes [13
]. A multi-country fluid intake survey confirmed that France was characterised by a high contribution of drinking water to total fluid intake [16
]. However, a study based on national population-based data of 2005–2007 estimated a TWI at 2285 mL/day for French adults aged 18–79 years old [13
], suggesting that a part of the French population is at risk of inadequate intake. Considering that in 2006–2007 about 90% of children aged four to 13 years in France failed to meet the EFSA water intake recommendations [17
], it seemed opportune to investigate adherence to EFSA guidelines among French adults.
Addressing the complex delineation of the role of fluids in a healthy diet, the US-led publication the Beverage Guidance Panel suggested that the consumption of water and other beverages with no or few calories should take precedence over the consumption of beverages with more calories [18
]. Further studies conducted on the US population found that an elevated consumption of drinking water—tap water and bottled water—was associated with higher nutritional quality, defined either by a healthier dietary pattern (i.e., greater consumption of vegetables, low-fat dairy products, and/or whole grains) [19
], a higher food variety [20
], the Healthy Eating Index (HEI) [20
], a better micronutrient adequacy [22
], or reduced energy intakes [20
]. An elevated consumption of drinking water was also associated with higher levels of physical activity [22
]. However, in France, there is a paucity of studies describing drinking water patterns in light of socio-demographic determinants and a complete lack of research on the association between drinking water intake and diet quality.
Based on data from a representative sample of the French adult population, the present study examined if there was an association between drinking water intake and diet quality assessed by several dietary indices. We hypothesised that the largest drinking water consumers had a better diet quality. We also estimated the adherence of men and women to the AI proposed by the EFSA and analysed the contributions of the different TWI sources.
2. Materials and Methods
2.1. Study Population
The second Individual and National Food Consumption Survey (INCA2) was carried out by ANSES (the French Agency for Food, Environmental, and Occupational Health) between December 2005 and May 2007 among representative samples of French adults and children to collect information on habitual food and beverage consumption. The samples were obtained using a multi-stage cluster sampling technique, established by the National Institute for Statistics and Economic Studies (INSEE). The sampling frame was approved by the French National Commission for Computed Data and Individual Freedom (Commission Nationale de l’Informatique et des Libertés, CNIL). The present analyses used data from the adult sample of the INCA2 (n
= 1918) including men (n
= 776) and women (n
= 1142) aged 18–79 years old. INCA2 remains the most recent version of a population-based survey available in France providing dietary intake information. A detailed survey methodology is available elsewhere [13
2.2. Demographic, Socio-Economic and Behavioural Variables
Individual socio-economic variables were collected using a self-reported questionnaire and an interview. The following information was available: sex, age, socio-occupational status, family status, education level, income per consumption unit, food insecurity, perception of household financial situation, educational level, residency, season of protocol completion, physical activity, and smoking status. A detailed description of these variables is available elsewhere [25
Socio-occupational status was classified into four categories: ‘low’, ‘intermediate’, ‘high’, and ‘economically inactive’. ‘High’ was assigned to executive, top-management, and professional classes; ‘intermediate’ to middle professions (office employees, technicians, and similar); and ‘low’ to manual workers and unemployed people. The fourth class, labelled as ‘economically inactive’, included retired people, students, and housewives/househusbands.
Family status was divided into ‘couples with children’, ‘couples without children’, ‘single parent households’, and ‘single without children’.
Education level was divided into ‘high’, ‘intermediate’, and ‘low’. ‘High’ was assigned to university education; ‘intermediate’ to high school; and ‘low’ to mid-secondary or below [26
Income per consumption unit (ICU) was calculated as the self-reported household total net income divided by the number of consumption units in the household. The number of consumption units was calculated using the Organization for Economic Co-operation and Development (OECD) modified equivalent scale (one consumption unit for the householder, 0.5 for other household members aged 14 or over and 0.3 to each child aged less than 14 years old) [27
]. For the analysis, the ICU was transformed into quintiles according to sex.
Food insecurity was classified into ‘yes’ or ‘no’ based on the perception of the actual situation in the household about having enough food or not, and the reason for a lack of food [25
]. Individuals having reported ‘getting enough, but not always the kinds of food they want to eat’, or ‘sometimes’ or ‘often not getting enough to eat’ because of ‘lack of money’ were classified as living in a household experiencing food insecurity.
The perception of household financial situation (‘living comfortably’, ‘getting by’, ‘finding it difficult’, ‘impossible without debt’) was assessed [28
] and further aggregated into two classes: ‘high’ and ‘low’.
Residency was recorded based on eight different regions of France: Northwest; East; Ile de France; West; Centre; Centre-East; Southwest; and Southeast.
Level of physical activity was based on the International Physical Activity Questionnaire (IPAQ) score [29
], which assesses physical activity, including exercise, leisure time, domestic and gardening activities, work-related and transport-related activity.
2.3. Dietary Assessment
Diet was assessed using a seven-day open-ended food record. Each day of the food record was divided into three main meals (breakfast, lunch, and dinner) and three between-meals snacks. The individuals were asked to describe, as precisely as possible, all food and beverage intakes for seven consecutive days: food name, origin (home-made or industrial product), and features (low fat, low sugar, fortified, dietetic, as well as fresh, canned, or frozen). Portion sizes were expressed by weight or household measures (spoon) or estimated using a photographic booklet (SU.VI.MAX) [30
]. Average daily nutritional intakes (excluding all alcoholic beverages) were evaluated matching food intakes with the 2013 French food composition database of the CIQUAL led by ANSES [31
]. Daily TWI, expressed in grams per day, was estimated by assessing the amount of the nutrient “water”, from fluids and from food moisture. Energy and quantity of alcoholic beverages were estimated separately.
2.4. Foods and Fluids Categorization
All the foods and fluids declared as consumed in INCA2 were categorised into nine food groups and 27 food subgroups (Supplementary Materials Table S1
). In our study, the food group “drinking water” contained tap water, and non-carbonated and carbonated non-caloric bottled water. The food group “beverages” included fruit juices, hot drinks, sugar-sweetened beverages, and diet sweet beverages. “Fluids” referred to the food groups “drinking water”, “beverages”, and the “milk” subgroup.
2.5. Nutritional Quality of Diet
Solid energy density (SED), mean adequacy ratio (MAR), mean excess ratio (MER), probability of adequate intakes (PANDiet) scores, and food variety were used as indicators of nutritional quality for each individual diet. The MAR, the PANDiet, and the MER are reliable indicators of the nutritional quality of diets at the population or individual level [32
]. SED (kcal/100 g) was defined as the ratio of the total energy consumed from solid foods and the total weight consumed from solid foods [35
]. A low SED diet has been associated with a good overall nutritional quality [35
The MAR (% of adequacy) was used as an indicator of good nutritional quality and was calculated for each individual diet as the mean percentage of sex- and age-specific French Recommended Dietary Allowances (RDA) [32
] for 23 key nutrients [37
]. The MER (% of excess) was calculated as the mean daily percentage of the French maximum recommended values for saturated fatty acids (22.2 g), free sugars (50 g), and sodium (3153 mg), as proposed by Vieux [34
]. The MAR and MER values range between 0% and 100%.
The PANDiet score was composed of adequacy probabilities for 24 nutrients grouped into two sub-scores: the adequacy sub-score (AS) and moderation sub-score (MS) [32
]. The AS assessed the probability of adequacy for items for which the usual intake should be above a reference value, whereas the MS evaluated the probability of adequacy for several items—recently adapted and including free sugar—for which the usual intake should not exceed a reference value [38
]. PANDiet scores range between 0 and 100; where 100 represents 100% of the usual intake adequacy for the 24 nutrients.
Food variety was estimated as the number of different foods and fluids (except alcohol) declared as consumed by each individual during the seven-day food record.
2.6. Diet Cost
Diet cost was calculated by multiplying the quantity of each food in the diet by its mean national price. A detailed methodology has been previously described [39
]. Diet cost was expressed either per day or per 2000 kcal (i.e., energy cost). Mean national prices, expressed in euros per 100 g of edible food, were previously obtained from the 2006 Kantar-World Panel database, which gives the annual food expenditures of a representative sample of 12,000 French households [40
2.7. Statistical Analysis
All analyses accounted for the complex INCA2 sampling frame design [24
]. Data were weighted for unequal sampling probabilities and for differential non-responses by region, agglomeration size, age, sex, occupation of the household head, size of the household and season [13
]. All analyses were conducted separately by sex due to the sex-specific EFSA AI of TWI (>2.5 L for men and >2 L for women). Small, medium, and large consumers were identified based on the tertiles of drinking water intake (including non-consumers). Socio-demographic characteristics were described and statistically compared between tertiles of drinking water intake using the chi-squared test (for qualitative variables) and general linear models (GLM, for continuous variables). Water intakes from fluids and food moisture, expressed in g/day and in percentage of TWI, were evaluated and represented graphically. The prevalence of adherence to the AI of TWI was assessed by tertile of drinking water intake and by sex using binomial logistic regression. The distance between TWI and the AI (i.e., TWI shortfall) for individuals considered in inadequacy was assessed. Distribution of TWI and the average TWI shortfall were graphically represented by tertile of water consumption by sex, and compared two by two using GLM, with Bonferroni correction.
Food intakes, food variety, SED, MAR (%), MER (%), PANDiet, diet cost (€/day), energy cost (€/2000 kcal), and macro- and micronutrient intakes (those used in the MAR or PANDiet) were statistically compared using GLM according to tertiles of drinking water intake in observed diets. Linear trends in diet quality and food intakes were also evaluated by tertile and by sex.
A p-value of 5% was used as the threshold of significance. Values are survey-weighted means and adjusted for total energy intake. When specified, adjustments were made for the level of education, socio-occupational status, season, level of physical activity, smoking status, region of residence, and quintile of ICU. Based upon the weighting factors, all results are representative of the French population. Analyses were conducted using SAS version 9.4 (SAS Institute, Cary, NC, USA).
This study was the first representative study of French adults investigating the associations by sex between drinking water patterns and diet quality in light of socio-economic determinants and adherence to EFSA AI for TWI. Our findings confirmed our hypothesis that an elevated drinking water intake was positively associated with diet quality, as large drinking water consumers were more likely to adhere to the EFSA AI, had diets of higher nutritional quality and, mostly among women, seemed to make healthier food choices (e.g., more fruits and vegetables and fewer sweets). Seventy-two percent of men and 46% of women in the French adult population were below the EFSA AI for TWI. In our sample, large drinking water consumers were more likely to have a high socio-occupational status and, among men only, to have a higher education level, a higher physical activity level, a higher income per consumption unit, and were more likely to be non-smokers.
It stems from our results that a higher drinking water intake was associated with higher nutritional quality of the diet, assessed by several dietary indices. In our study, the MAR and the PANDiet were positively associated with drinking water while differing in their methods, as the MAR is a simple mean percentage of sex- and age-specific French RDA [37
], whereas the PANDiet is a score that takes into account different parameters, including the number of days of dietary data, the mean nutrient intake and its day-to-day variability, the nutrient reference value, and inter-individual variability [38
]. Using different indices is useful to show that our observations are consistent and do not depend on a certain methodology of assessing diet quality. All relationships between tertiles of drinking water and dietary indices, SED, and energy cost, even though not all were significant, were congruent in the same direction. Full adjustment revealed that the association between drinking water intake and diet quality was particularly noticeable for women, for whom drinking water increased, as did the indicators of nutritional quality (i.e., higher MAR and PANDiet scores, and lower SED values). Results are consistent with Kim et al., who found a positive association, among both Korean men and women, between drinking water intake and the MAR [41
]. When not using dietary indices, other relationships between drinking water intake and nutrients have been found in the literature that we could consider as good indicators of nutritional quality. In the US, Yang and Chun found a positive association between TWI, drinking water, and moisture in foods with dietary and serum minerals, vitamins, and carotenoids [22
]. Those nutrients positively associated with TWI can be compared to identical components of the MAR, and are in line with our results. In a recent national study of US adults, an increase in the proportion of daily drinking water in TWI was found to be associated with a decreased daily intake of total energy, energy from sugar-sweetened beverages, discretionary foods, and total fat, saturated fat, sugar, sodium, and cholesterol [21
]. Most of those nutrients, the intake of which is negatively associated with drinking water, are similar to the components of the MER, and are also slightly decreased in our results among men. Higher intakes of sodium found in women could possibly explain why the MER did not significantly decrease. In previous studies, a positive association was found between drinking water and sodium intakes [4
], similar to our results among women, although the cause-consequence relationship is unknown.
In terms of fluid patterns, our results indicate that large drinking water consumers seemed to favour a higher intake of only water and not necessarily all other fluids. A similar observation was made by Illescas-Zarate et al. in her study among Mexican adults, in which a negative association was found between drinking water intake and sugar-sweetened beverages [42
In terms of food choices, large drinking water consumers seemed to favour diets rich in moisture-abundant foods, especially fresh fruits. This finding is in line with the literature reporting the opposite observation, i.e., low levels of fruit and vegetable intake among low drinking water consumers [23
]. Similarly, a negative association between drinking water intake and beverage moisture, but a positive association with food moisture, was found among US adults [20
]. Kant et al. suggested some substitution effect of drinking water intake on other fluid consumption and possibly higher fruit and vegetable intake [20
]. According to Hedrick et al., consumption of water, unsweetened tea/coffee, low-fat milk, artificially sweetened beverages, and fruit/vegetable juice is closely aligned with a “prudent” dietary pattern (usually including vegetables, fruits, legumes, whole grains, fish, and poultry); conversely, the consumption of high-fat milk, alcohol, and sugar-sweetened beverages is strongly associated with a “Western” dietary pattern (usually including red meat, processed meat, refined grains, sweets and dessert, French fries, and high-fat dairy products) [43
]. This last point is particularly illustrated in our results with the decrease of mixed dishes and sandwiches among men and of sweet products among women from the first to third tertile of drinking water intake. Large drinking water consumers tended to consume more moisture-abundant foods, also rich in nutrients and vitamins, implying that people fulfilling their water intake requirements have higher quality diets and make food choices more likely to fulfil their nutritional requirements.
Identifying characteristics of large drinking water consumers can be useful to identify individuals more likely to have healthier dietary behaviours. In the literature, associations have been found between drinking water intake and some socioeconomic determinants, such as sex [42
], age [22
], education level [22
], income [23
], level of physical activity [22
], and smoking status [11
]. The novelty of our study is that drinking water patterns were explored in light of several factors combined (socio-demographics, lifestyle determinants, and overall dietary intake). Moreover, a major strength of our study is the investigation of socioeconomic determinants of drinking water intake by sex in order to identify gender specificities, which has only been done before, to our knowledge, by Kim et al. [41
]. In our study, both among men and women, drinking water intake was positively associated with socioeconomic position. However, female drinking water intake appears to be less influenced by socioeconomic factors, considering the numerous additional associations for men, notably with education, physical activity, income, and smoking status. Thus, among men, drinking water intake may be considered as a reliable indicator of socioeconomic differences. This male specificity has not been noticed by Kim et al. among Korean adults [41
]. Among men, the association between drinking water intake and diet quality was explained by their socio-economic status, as men in higher socio-economic strata had a higher nutritional quality and higher drinking water intake. However, among women, the association between drinking water intake and a healthy dietary pattern was independent of the social level. Gender contrasts in food choices are influenced by views on food and health, the ethical dimensions of food production and food selection, nutritional attitudes and choices, dietary change, food work, and body image [46
]. One hypothesis could be that, regarding beverages, women appear to be more ‘health-conscious’ [47
] than males, who consume more alcohol [14
] and sweet beverages regularly [14
]. More research is needed to fully apprehend divergent associations found in the literature and to verify if women and men follow a distinct drinking water pattern elsewhere than in France.
In the current study, daily means of TWI were consistent with results from previous European studies [11
]. However, a remarkable observation was made in that the sex difference in TWI was less pronounced in France than in some other countries: estimated TWI was slightly lower for men and slightly higher for women than those of Irish [11
], British [14
], and German [50
] adults (2.16 L vs. 2.52 L, 2.53 L, and 2.48 L for men, and 2.12 L vs. 2.09 L, 2.03 L, and 2.05 L for women, respectively). This is, on one hand, in contrast to a tendency reported in the literature with men systematically having a higher TWI than women [4
] while, on the other hand, fluid surveys in 13 countries showed few significant differences in fluid intake between both sexes. The latter suggests that if there is a gender difference in TWI, the difference could be due to a difference in water from food moisture, a result that was also observed in our study, but needs to be confirmed further.
Several strengths and limitations of this study should be acknowledged. The INCA2 data were collected in 2006–2007 and the observations may already be obsolete. However, INCA2 is still the most recent version of a population-based survey available in France and remains the standard source of reliable information about dietary intakes. Furthermore, between the first version of the national survey (INCA1) and the second (INCA2), the food group ‘waters’ increased from six items to 50 items, increasing the robustness of data from this food group, which is of particular importance for the present study [51
]. A potential additional limitation could come from numerous low drinking water consumers interviewed during the winter. However, no interaction effect was found between the season and drinking consumption patterns on the level of adequacy (data not shown). Another limitation, inherent to any dietary survey, is the fact that the nutritional intake data is self-reported. Especially collecting data to evaluate TWI is not without limitations: accurate recording of drinking water and other beverages, as well as estimating water from food moisture, might be prone to bias [52
]. A final point of discussion is the exclusion of alcohol. Since several publications report a major contribution of alcoholic beverages to water intake [11
], excluding these beverages could lead to an overestimation of the proportion of French adults below the EFSA AI of TWI. However, this could also be interpreted as a limitation in our case, since having an adequate intake of TWI should be achieved mainly with drinking water.