Calculation method for daily water intake components using doubly labeled water

Daily water intake (DWI) is essential for survival in humans; however, accurate assessment of DWI from drinks and beverages (Wdrinks) or food moisture (Wfoods) is dicult as it depends on self-reported intakes that are prone to inaccuracy. Here, we established an objective method to assess DWI components using doubly labeled water (DLW). Deuterium and H218O were orally administered, and the dilution space and elimination rate of 2H and 18O were measured. DWI was calculated from the deuterium turnover corrected for metabolic water production and insensible water absorption from humidity. Wfoods was estimated using dietary record (Wfoods-DR) or calculated from the total energy expenditure assessed by DLW (Wfoods-DLW). The current results underscore Wfoods-DR underestimation using self-reported dietary assessments, which underestimates food intake. This study proposes novel methods for calculating each DWI component using DLW.


Introduction
Water, a critical but overlooked nutrient 1 , is the principal chemical constituent of the human body, accounting for approximately 65% and 50% of body weight in infants and elderly persons, respectively [2][3][4] . Water lls intra-and extracellular compartments, and it serves as a solvent for minerals, vitamins, amino acids, glucose, and many other nutrients. Thermoregulatory, circulatory, and urinary systems are dependent on body water 5 and water balance; water loss and intake are regulated by peripheral and central homeostatic systems 6 . Observational studies suggest that adequate hydration and/or prevention of acute dehydration are associated with reduced risks of developing various morbidities [7][8][9][10] .
Although epidemiological studies have attempted to examine the effects of daily water intake (DWI) on health outcomes, the accumulation of scienti c evidence is delayed 8,11 . A major limitation of the abovementioned studies is that estimated DWI is based on self-reported dietary assessment methods.
Results of multi-country comparison studies suggest that the amount of beverage intake depends heavily on a variety of dietary assessment methods, not only on sampling bias or cultural differences 12 . In addition, total water in ux does not originate only from Wdrinks but also from Wfoods, metabolic water production (Wmet), respiratory water uptake (Wres), and transcutaneous water uptake (Wtra) (Fig. 1).
Restricted food intake enhances drinking behaviors, 13 and metabolic water plays an important role in animals living in environments with limited access to drinking water and food moisture 14 . Thus, to examine the relationship between DWI and health outcomes, it is necessary to assess each source of DWI accurately.
What percentage of DWI originates from Wdrinks or Wfoods? The Third National Health and Nutrition Examination Survey in the United States of America (USA) indicated that approximately 80% of DWI is obtained from drinks and beverages and 20% from food. Moreover, other studies also suggested that 20-30% of DWI originates from food moisture [15][16][17] . However, subjective dietary assessment methods underestimate 10-40% of energy and protein intakes compared to assessment with biomarkers 18,19 .
This fact suggests that subjective dietary assessments underestimate total food intake, resulting in underestimation of Wfoods as well. However, to the best of our knowledge, no previous studies have attempted to calibrate Wfoods using biomarkers.
Several approaches have been proposed for the biomarker calibrations of energy, protein [20][21][22][23] , and intake of various minerals 24 . The doubly labeled water (DLW) technique is considered a criterion method for energy intake (EI) calibration. Previous studies indicated that the calibrations make associations between nutrient intakes and health outcomes clear 25,26 . We report here a method for assessing DWI components, particularly differentiating Wfoods from Wdrinks, using DLW. We hypothesized that DWI from food moisture accounts for a much higher percentage of total DWI than presented in current consensus, guidelines, or literature. If this hypothesis is true, the importance of food moisture should be considered in dietary references for water intake and/or water intake advices in public health and clinical settings. Epidemiological studies that examine the health effects of DWI should be conducted to evaluate DWI from both foods and liquids with adequate biomarker calibration using DLW.

Results
The physical characteristics and body composition of the 141 participants are shown in Table 1. Daily water turnover (rH 2 O), energy expenditure, and body composition were measured using the DLW method, as previously described 21 The elimination rates of 18 O and 2 H (ko and kd, respectively) were determined, and carbon dioxide production rate (rCO 2 ) (mol/d) was calculated with the assumption that isotope fractionation applies to water in breath and transcutaneous water using Eq. A6 by Schoeller et al. 33 ; the revised dilution space constant was provided by Sagayama et al.. The rCO 2 (L/d) was obtained using the following formula: We assumed that the respiratory quotient (RQ) was 0.87 based on the food quotient 34 , and total energy expenditure (TEE) was calculated using the modi ed Weir's equation as follows: TEE = 1.106 × rCO 2 + 3.94 × (rCO 2 /RQ) × 4.184 / 10 3 [3] The quality checklist is described in the International Atomic Energy Agency documents 30 .

Resources of daily water intake
The daily total water turnover (TWT, L/d) was calculated using the DLW method according to ref. 35 Eq. 5 as follows: where f is hydrogen isotope fractionation. When body water is maintained, rH 2 O is equal to water e ux and in ux (i.e. DWI). Metabolic water production (Wmet; L/d) was calculated from energy expenditure and from assuming a metabolic mixture of carbohydrate, fat, and protein in a caloric ratio based on the ref. 35 Eq. as follows: Wmet = TEE × [(%fat × 0.119) + (%protein × 0.103) + (%carbohydrates × 0.15) + (%alcohol × 0.168)] [5] Inspiratory water (Wins; L/d) was calculated as: Wins = respiratory air volume × estimated absolute humidity / 1000 [6] where respiratory volume is in L/d and absolute humidity is in mg/L, estimated from predicted air temperature. Respiratory air volume was calculated from rCO 2 obtained from DLW, assuming that 3.5% of expired air is CO 2 .
Transcutaneous water in ux (Wtra) was then calculated as 27 : Wtra = 0.18 × (absolute humidity/21.7) × 0.5 × BSA × 1.44, [7] where 0.18 is the rate of transcutaneous absorption in g/m 2 of body surface area (BSA) in an atmosphere saturated with water vapor (21.7 mg/L). The BSA (m 2 ) was estimated using the Dubois formula, and a clothing factor of 50% was assumed, as clothing would decrease the rate of evaporation through the skin.
Dietary water intake from preformed water (DWI) was calculated as the difference between rH 2 O and the sum of all the above-calculated values (Wmet, Wins, and Wtra) as follows 27 : Wfoods was obtained using the following equation: where the water content of food (WCF) and energy density (ED) are obtained in L/kg and MJ/kg, respectively. Wdrinks was calculated as the difference between DWI and Wfoods. TWT and its sources in the participants are shown in Table 2. Although Wfood-DR was signi cantly correlated with total EI estimated using DR (R 2 = 0.191, P<0.001), Wfood-DR was not correlated with TEE (R 2 = 0.000 P=0.856, Figure 2A). Wfood-DR was signi cantly lower than Wfood-DLW (P<0.001, Bland-Altman plot, Table 2 and Figure 2C). These data indicated that DRs could not be used to accurately and precisely assess DWI from food.
EI assessed using DR was signi cantly lower than TEE (P<0.001, see Bland-Altman plot, Figure 2C). Figure 2D shows the relationship of the percent differences between EI (by DR) and TEE, with the percent differences between Wfood-DR and Wfood-DLW. The percent difference between Wfood-DR and Wfood-DLW was proportionally associated with that between EI by DR and TEE (R 2 = 0.93, P<0.001). The estimation error of Wfood-DR is dependent on the estimation error of EI by DR. We applied Eq. 11 in the validated group (n = 69). In Figure 3B, the black line is the y = x line, and the intercept and slope of the regression line are not signi cantly different from those of the y = x line.
Therefore, the equation was validated. We pooled all participants (n = 141) and got the regression line as follows ( Figure 3C): Wfood (L/day) = 0.0914 × TEE + 0.315 [12] After calculation, the proportions of the component of DWI, Wfood and W uid were 53% and 47%, respectively.

Discussion
First, we discovered that self-report dietary assessment methods systematically underestimate Wfood. The underestimation of Wfood is highly correlated with the underestimation of EI by DR when compared to TEE by DLW. In addition, self-reported Wfood-DR was not signi cantly correlated with Wfood-DLW. These results suggest that reliable methods are required to accurately and precisely assess Wfood. The newly established equation can reasonably estimate Wfood. This is because Wmet, Wres, Wtra, and Wfood can be assessed using the DLW method, and W uid can be estimated as the residual of them from TWT.
It has been generally considered that 70-80% of DWI is obtained from uids (drinks and beverages) and only 20-30% from foods [15][16][17] . However, the present study indicated that approximately 45% and 55% of DWI are from foods and drinks/beverages, respectively. DWI from food accounts for a much higher percentage of total DWI than is revealed by current consensus, guidelines, or literature. This is because dietary assessment methods underestimate DWI from food as well as energy and protein intake, although no previous studies have demonstrated this underestimation. TWT can be measured using the deuterium pool size and the washout, and systematic underestimation of DWI from food induces systematic overestimation of DWI from drinks. Our results suggest that the current guideline for drinking water and beverages should be revised based on the biomarker method. The current guideline of the Institute of Medicine in the USA 15 considered only 0.5-0.7 L/d as Wfood. Our participants actually had 1.12 L/d of Wfood, which is almost double the value in the current consensus, although the weight and height of the current participants are lesser than those of the normal US population. DWI in the current consensus led to a recommendation of in ated beverage consumption. In summary, our results emphasized Wfood as an overlooked water source. Moreover, these results underscore the importance of DWI assessment using objective methods and the need for a revision of current guidelines in terms of drinking water and beverages. Bland, J.

Study cohorts
This study analyzed subgroups of the Kyoto-Kameoka Study wherein DLW was measured. Details of the abovementioned study have been previously described 21,37-39 . We sent mail invitations for face-to-face physical examination to 4831 older adults aged at least 65 years, who lived in Kameoka city, Kyoto, and had responded to a baseline mail survey in February 2012. Of the invited individuals, 1379 participated in a face-to-face physical checkup between March and April, with a participation rate of 30.3%. We advertised the opportunity for individuals to have their energy expenditure measured using DLW, and 147 individuals voluntarily participated in the measurement between May and June 2012. We excluded participants who had missing data in a 7-day dietary record (DR) or DLW (3 people each); therefore, data from 141 individuals (62 women and 77 men) were analyzed. Participants were divided into a model development group (n = 72) and a validated group (n = 69), using the R software function of random number generation.

Dietary records
We used a previously described protocol for dietary assessment 21 . We collected DRs over 7 consecutive days during the DLW method to include both weekdays and weekends. During an informational meeting, the research staff (registered dietitians) were educated on how to administer the DR to the participants, using completed DR sheets as examples. Each participant was provided blank record sheets, wherein to provide their DRs, as well as a digital scale (TANITA, Tokyo, Japan) and paper media for education.
Research dietitians instructed the participants to record every food item and beverage consumed daily during or between meals.
The dietitians checked all completed records at each participant's home and reviewed them at least twice in a standardized manner. Research dietitians coded and entered completed DRs into an energy and nutrient analysis program called WELLNESS21 software (TopBusinessSystem, Okayama, Japan), which conforms to the Standard Tables of Food Composition in Japan. Participant-recorded foods that were not listed in the Standard Tables of Food Composition in Japan were replaced with similar foods. The DR was used to calculate individual values of EI, WCF (%), ED (MJ/day), and dietary water intake from food (Wfood-DR).

Ethics
All participants were informed of the purpose, procedures, and risks of the study, after which they provided written informed consent before participation. The ethics committee of Kyoto Prefectural University of Medicine and the National Institute of Health and Nutrition approved the study protocol (RBMR-E-372 and NIHN187-3). The current analysis used baseline data and did not contain any intervention parts of the study, but the Kyoto-Kameoka study included intervention studies with data on physical activity and nutrition promotion. Therefore, this study has been registered in a clinical trial database (UMIN000008105). The research protocols, outcomes, and measurements including DLW and dietary assessments were fully described at research protocol papers 39,40 . Declarations YY, DW, HS, HP, MM, JRS, DAS established the concept of this methodology; YY, MK designed research of cohort; YY, HS, AI, TY, YW, HN, EY, MK conducted the study; YY, DW, HS analyzed data; YY wrote the manuscript; AL, HP, JRS, DAS revised the draft substantially; YY had primary responsibility for the nal content. All authors read and approved the nal manuscript.

Con icts of interest
None of the authors has con icts of interest.   5.49 ± 0.74 TEE, total energy expenditure; EI, energy intake; TWT, total water turnover; Wmet, metabolic water production; Wres, respiratory water uptake; Wtra, transcutaneous water uptake; DWI, daily water intake, Wfluid, preformed water in ingested liquids (drinking water and beverages); Wfood, preformed water in ingested foods; Wfood-DLW, Wfood estimated by DLW only, Wfood-DR, Wfood estimated by DR only; WCF, water content of food; ED, energy density. *** EI assessed by DR was significantly lower than TEE measured by the DLW method (P<0.001). ### Wfood assessed by DR was significantly lower than Wfood measured by the DLW method. (P<0.001). Figure 1 Components of total water turnover (TWT). Daily water intake (DWI) can be separated into DWI from ingested liquids (Wdrinks) and from ingested foods (Wfoods) Figure 2 Relationship between Wfoods-DR and TEE (A). Wfood-DR was not correlated with TEE. Bland-Altman plot shows the DR underestimated actual Wfoods (B), which is consistent of the underestimation of energy intake by DR (C). The estimation error of Wfood-DR is dependent with estimation error of EI by DR (D).

Figure 3
Relationship between Wfoods and TEE in the model developed group (n = 72) (a). We applied the developed equation in the validated group (n = 69) (b). The intercept and slope of the regression line are not signi cantly different from those of the y = x line. We pooled all participants (n = 141) and got the regression line (c).