Using 12068 days of weighed dietary records from a nationally-representative sample of British adults aged 19 to 64 years, we have quantified total water intakes, and investigated relationships between beverage consumption patterns and water and energy intakes.
Mean TWI was almost identical to the European (EFSA) reference or “adequate intake” (AI) of 2 L for women and 2.5 L for men . EFSA determined their AIs from a combination of observed intakes in population groups, with desirable urine osmolarity values and desirable water volumes per energy unit consumed. These AIs only apply to conditions of moderate environmental temperature and moderate physical activity levels; physically active individuals or those working in hot conditions may require more. However, AIs are likely to exceed requirements and cannot predict which individuals are under- or over- hydrated in practice. Using more conservative criteria (TWI below AI and TWI <1 g per kcal of energy intake) approximately 33% of men and 23% of women were classified as having low water intakes. These may be at greater theoretical risk of poor hydration, but may actually be appropriately hydrated if they have low volumes of water loss due to their individual physiology, environment and physical activity. In our opinion a value of 1 g water per kcal energy requirement (rather than energy intake) would seem an appropriate basis for a UK recommended mean water intake, and has the utility of broad applicability across age/sex groups. However, it would be helpful to attempt to specify a confidence interval around this mean to avoid confusion and misuse.
Beverages supplied 75% of TWI for adults in this British study. This is consistent with other estimates [4, 7, 8], but the proportion may vary considerably between individuals and between populations. As we have shown, beverage consumption (and water intake) is not evenly spread throughout the day, but tends to be concentrated in the evening. This leaves the possibility that some adults may be relatively overhydrated in the evening/nighttime and under-hydrated in the morning. As alcoholic beverages were a significant source of water and energy, especially among men, and were strongly associated with evening consumption, men who drink alcohol may be particularly vulnerable. TWI can be increased by offering more variety in beverages . However, if such an intervention is to benefit public health, it is important that the additional fluid is consumed when needed, and does not result in excessive water intake, unwanted effects on energy balance, or inadequate nutrition.
Since our analysis was conducted, Ng et al.  have reported on UK trends in beverage consumption over the past 25 years. There are some differences in methodology between our studies, but their calculation for adults in 2000/2001 (18% of energy obtained from beverages) compares with our estimate of 16% (which excluded discretionary sugar). Had we been able to include table sugar added to drinks, we would have seen a small contribution of tea and coffee to energy intake but our overall findings would not have been affected. Between 2000 and 2008/9 the energy intake contributed by beverages (in total) did not change, while sweetened tea and coffee declined and consumption of alcoholic beverages and caloric soft drinks rose, although not significantly . In 2008/9 alcoholic beverages provided 770 kJ/d (184 kcal/d) per capita, and caloric soft drinks (soda and juice drinks) 209 kJ (50 kcal/d) per capita; hence our conclusions concerning the dominance of alcohol appear to remain valid. According to Ng et al., TWI increased by approximately 9% between 2000/2001 and 2008/2009, as people consumed more water as a beverage and more water in food (fruit and vegetables) , a trend that may have been influenced by healthy eating campaigns and marketing of bottled water. However, consumption of water, tea and coffee, soft drinks and juice among adults has changed little over the last 3 years (NDNS 2008–2011) [27, 28], and plain water consumption in the UK remains lower than among adults in France , Canada  or America . Consumption habits are particularly well- documented in the US, and NHANES surveys show alcohol providing (only) 115 kcal/d for adults, while soda /fruit drinks provided 141kcaL/d in 2005/2006 . It appears that the main increase in soft drink consumption in the US occurred between the 1970s and the 1990s, and a new study confirms that liquid calories have been falling in the US diet over the past decade, although consumption remains high .
The present study demonstrates that well-conducted national surveys such as the NDNS have the potential to yield rich contextual data that can be linked with nutrition and health measures. We found significant variation in consumption by day of the week and by time of day (see Additional file 1: Appendix for supplementary figures), which few other studies have been able to describe in detail. Alcoholic drinks in particular were a major contributor to evening and weekend peaks in beverage consumption. The extent to which timing of drinking occasions relates to overall TWI and EI requires further study. Associations have also been demonstrated between meal patterns and EI, with the suggestion that evening consumption of calories may be more conducive to total energy excess . However these findings may merely reflect the cultural reality of the evening being a time for eating and drinking. Significantly, and possibly for the first time using such data, we have estimated the effect on EI of switching beverage consumption between caloric and non-caloric sources using within-person daily records.
In terms of the limitations of this study we draw attention to the usual caveats on making causal inferences from observational data. Whilst the data are of the highest quality obtainable, and respondents were revisited to check records and probe for missing items and weights, dietary assessment is an imperfect science. Most errors are likely to be in the direction of omission or underestimation (perhaps especially where alcohol is concerned). Errors of overestimation may also have occurred where a drink was spilt or unfinished, although efforts were made to prevent this by asking respondents to weigh leftovers and estimate spills. There is evidence from validation studies that true EI is underestimated by about 20-25% in the NDNS , but to our knowledge no studies have tried to quantify under-reporting of water intake or beverages specifically. Estimates of water intake adequacy based on survey data are likely to suffer from under-estimation also.
In conclusion, a significant proportion of British adults surveyed in 2000/2001 had low TWI, although in the absence of clinical measures this could not be equated with poor hydration status. Further work may be warranted to assess current intakes and to explore correlations with urine volumes, as recommended recently . Soft drinks were not a major source of liquid calories among these British adults, whereas alcohol was more significant. All beverages supplying energy can contribute to higher total EI, but in practice some degree of compensatory under-consumption of other foods or beverages may occur. Mechanistic and experimental studies are required to address directly the satiety effects of beverages, while work is also needed to understand the drivers of beverage consumption, which are not merely physiological but also psychological, social and environmental. Further research based on dietary patterns rather than singular foods or dietary components may help identify unhealthy behaviors and provide better evidenced-based recommendations for adequate fluid intake and optimal beverage consumption.