Introduction
Low fluid intake, or its biomarkers, has been associated with an increased risk of developing cardiometabolic diseases [
1,
2], chronic kidney disease [
3] and recurrent kidney stones [
4]. Adequate hydration has been shown to improve cognition in children [
5] and adults [
6], improve mood [
7,
8] and attenuate biological risk factors for the above-mentioned conditions. In addition, consuming sugar-sweetened beverages (SSB) has been linked with an increased future risk of obesity [
9], and cardiometabolic diseases. Each daily serving increase in SSB has been linked to a 21% increased risk of type 2 diabetes [
10], 7% increased risk of hypertension [
11] and 7–18% increased risk of stroke, heart failure and coronary heart disease [
12].
Total fluid intake has been reported for many countries [
13‐
20] together with data on the type of fluid consumed [
21‐
25]. Many studies have shown that a high proportion of children and adolescents do not drink enough to meet water intake recommendations [
13,
15,
16,
23]. This is especially important in young children who, due to their relatively high body water content and underdeveloped regulatory systems, are vulnerable to the effects of not drinking enough [
26]. It is apparent from these, and many other studies, that while an individual may be drinking sufficient in terms of volume to meet, or exceed current recommendations on fluid intake, there may be a wide variety of combinations of fluid type within that total volume. Considering only one or two variables at a time limits the interpretation of data, and as a consequence, their usefulness [
27]. Increasingly, dietary patterns are being studied to investigate these interactions [
28]. Therefore, it is now pertinent to look at patterns of fluid consumption as opposed to studies that consider only volume or individual fluid type.
Different methods of analyzing dietary patterns can be used in diverse populations, including principal component analysis, cluster analyses and more recently, reduced rank regression [
29]. Fluid intake has been evaluated using multiple regression analysis [
19], principal component analysis [
30] and a dynamic panel model [
31]. Another multivariate technique that can be applied to fluid intake is cluster analysis. The main advantage of cluster analysis is that it creates groups of individuals that are as homogeneous as possible, minimizing the variance within each group and maximizing the variance between groups. This allows the identification of fluid intake patterns, called ‘clusters’, common to a group of people that are different from each of the other patterns. Such clusters can then be evaluated through classical regression methods, interpreted and validated from a fluid intake point of view. Cluster analysis has been used extensively to evaluate dietary patterns [
32], and is now being applied to fluid intake patterns in adults [
17,
33,
34] and children [
35‐
39], although its use is still limited.
However, to the best of our knowledge cluster analysis of fluid intake patterns in children and adolescents has only been performed in single-country population groups. A global analysis of FI patterns has not been possible to date due to variations in the methodology used to collect data; therefore, the importance of the country of residence per se has not been investigated using cluster analysis. With the availability of a validated 7-day fluid record [
40] and a harmonized methodology across various countries [
41‐
43] it is now possible to study global fluid intake patterns taking into consideration this variable. Therefore, the primary aim of this study was to identify different patterns of fluid intake in children and adolescents in six countries. The secondary aim was to characterize these patterns in terms of socio-demographics and lifestyle indicators.
Discussion
Cluster analysis enables the identification of behaviors and associated characteristics, which may help target the specific populations that require intervention to change behavior. Unlike other cluster analyses, that have looked at fluid intake types in the context of energy content [
35,
38], or diet quality [
36], the emphasis of this analysis was drinking behavior. A total of eight fluid types were used in this cluster analysis to identify fluid intake patterns in children and adolescents across six countries. The first observation was that the FI patterns in this analysis were driven by SSB and water, and to a lesser extent milk and its derivatives. The second most striking observation was that the most significant characteristic across the FI patterns was country of residence. The Latin American countries were more represented in the
low drinkers–SSB or the
high drinkers–SSB FI patterns. These results are consistent with reports based on volumes of fluid types consumed [
21,
42]. High intakes of SSB in Latin American countries have already raised concerns given the associations with dental caries [
50], obesity and overweight [
51] and associated metabolic conditions among children and adolescent [
52]. In addition to high SSB intake, residence in Brazil, Mexico and Argentina was associated with low total fluid intake (
low drinkers–SSB). The mean total fluid intake for this cluster was 1 L/day, which compares unfavorably with recommendations on the adequate intake of water from fluids [
53]. Children and adolescents in these countries may be at risk of suboptimal hydration [
54], which is associated with impaired cognition and low mood [
55] and physical performance [
56]. Uruguayan children and adolescents had high TFI and were over-represented in the
high drinkers–SSB pattern. Therefore, interventions may be most effective if targeted at replacing energy-containing drink with water while maintaining the TFI. This analysis suggests that interventions in these countries should be targeted at increasing drinking water consumption while reducing sugar fluid intake. Some countries, including Mexico [
57], are adopting this approach by introducing taxes to increase the price of sugar drinks; it is possible that this will particularly influence those in lower SELs [
58].
Chinese residents were predominantly in the
low drinkers–water and milk FI cluster (51%); a pattern in which the younger children (4–9 years) were over-represented; as were taking a lunch box to school and having fluid available at school. This pattern appears to reflect the policies implemented in China whereby a school lunch that includes a serving of 250 mL of milk is provided to first and middle-school children [
59,
60]. Indonesians were over-represented in both the
high drinkers– and
very high drinkers–water patterns; these FI patterns have TFIs > 2.5 L/day as shown in a previous analysis [
41]. From the present analysis, it would appear that in terms of fluid volume and type, there is little of concern that requires intervention in these latter two clusters. This could be due to all proactive actions undertaken in Indonesia to increase the access to safe water [
61] and to encourage water consumption [
62]. However, with increasing levels of obesity [
63] and type 2 diabetes [
64] in Indonesia there are concerns about increasing SSB consumption [
65]. Public health policies and interventions are needed to halt and hopefully reverse this trend [
66].
Fluid pattern analyses in children and adolescents have been conducted in the USA [
30] and Canada [
35]. Bougatsas et al. identified six clusters [
30] including one which was similar to the
low drinkers–water and milk cluster. Other comparisons between the two analyses are difficult due to the differences in fluid type classification. The analysis by Danyliw and colleagues [
35] identified five clusters including milk and high-fat milk; however, water was excluded from the analysis as fluids were categorized on the basis of their energy and nutrient content. Other analyses have concentrated entirely on energy content rather than volume of fluid types. This is the first study to include data from more than one country; therefore, direct comparisons to other analyses are difficult and reasons for any similarities or differences are beyond the remit of the present analysis. However, while societal and cultural influences on food patterns are well recognized [
67], identification of the factors that influence drinking patterns requires further research.
Socioeconomic level was a significant characteristic in three of the FI clusters. The
low drinkers–milk and water cluster was more represented by SES AB,
high drinkers–SSB were associated with SES DE and
very high drinkers–water were associated with SES C. Disparities in SES and hydration [
68], water [
69] and type of fluids consumed [
70] have been reported although evidence from cluster analysis is limited and complicated by the lack of consistent definitions of SEL. Danyliw et al. [
35] noted differences in intake according to household security and income, but no specific patterns were established. In a recent systematic review lower SES was associated with higher SSB consumption [
71]. In the present analysis, a lower SES was also a significant characteristic of the
high intake–SSB cluster. Therefore, it would seem appropriate for interventions in those countries that aim to decrease SSB consumption and replace SSB with water, target children and adolescents in the lower SELs.
Sedentary behavior and/or physical activity were significant characteristics of four of the FI pattern clusters. Participants who reported less than 2 h of sedentary behavior were over-represented in the
low drinkers–water and milk cluster. Children and adolescents who reported more than 2 h of sedentary behavior were over-represented in the
high drinkers with SSB cluster; those who reported being physically active once a week to twice a month were under-represented. A recent cluster analysis of data from the ELANA and HELENA studies in children and adolescents reported clusters characterized by sedentary behavior and SSB consumption [
72]. This is not surprising given the established links between sedentary behavior and less healthy dietary intake including SSB consumption [
71,
73,
74]. Those who reported sedentary behavior greater than 2 h a week and/or being physically active once a week to twice a month were over-represented in the
very high drinkers–water cluster. While this may appear contradictory, this phenomenon has been reported before and is probably due to those who report being sedentary for two or more hours per day compensating by being physically active at other times of the day [
72]. Clearly, the prevention and treatment of overweight and obesity in children and adolescents requires a multifaceted approach, which focuses on changing dietary habits, including reducing SSB consumption, and reducing sedentary behavior and or increasing physical activity.
The current study has several strengths including the use of a harmonized sampling and data collection methodology across the countries and of a validated assessment method for total fluid intake [
40], reflecting the day-to-day behavior of the participants over a 7-day period. The sample size and the use of data from participants in six countries undoubtedly strengthened the analysis. Objective statistical criteria such as using the silhouette coefficient to identify the number of clusters combined with the subsequent use of subjective criteria rendered the selected clusters interpretable. However, it is important to recognize the limitations of this study. Missing data for some of the variables resulted in a slightly reduced sample size; however, this is inevitable in large cross-sectional studies such as the Liq.In
7 survey. The use of biomarkers for hydration status or health outcome measures would have strengthened the findings and possible implications of the analysis. Parents or primary carers recorded fluid intake and responded to questions about lifestyle and socio-demographics for younger participants, while this may have increased precision they may have been biased towards demonstrated healthy characteristics. In addition, adolescents were not asked if their parents provided a lunchbox for school nor about water availability in schools. While the questions on sedentary behavior and physical activity provide vital and interesting information, it would have been better to have used a validated physical activity questionnaire such as the International Physical Activity Questionnaire (IPAQ). To ensure a reliable and sensitive approximation of socioeconomic status, country-specific methods were used as a harmonized classification system is not currently available.
This analysis is the first to investigate fluid intake patterns across countries and has shown that country of residence is an important determinant of cluster membership. Therefore, it would be interesting to repeat the analysis within each country and extend the survey to other countries and regions of the world. Given the interest in establishing guidance and recommendations across regions, e.g., Latin American countries, or continents, e.g., Europe, once more data are available it would be interesting to repeat the analysis again within these regions. Cluster analysis of fluid intake patterns could be a useful tool for monitoring interventions aimed at increasing water intake while reducing SSB consumption by repeating the analysis over a period of time.