- Email: [email protected]
The relationship of spontaneous macronutrient and sodium intake with fluid ingestion and thirst in humans
Physiology& Behavior, Vol. 49, pp. 513-519. ©PergamonPress plc, 1991. Printedin the U.S.A.
0031-9384/91 $3.00 + .00
The Relationship of Spontaneous Macronutrient and Sodium Intake With Fluid Ingestion and Thirst in Humans J O H N M. DE C A S T R O 1
Department of Psychology and the Behavior and Neurobiology Program Georgia State University, Atlanta, GA 30303-3083 Received 2 July 1990
DE CASTRO, J. M. The relationship of spontaneous macronutrient and sodium intake with fluid ingestion and thirst in humans. PHYSIOL BEHAV 49(3) 513-519, 1991.--The amount of solid food eaten by humans in spontaneously ingested bouts is the most important determinant of the amount and timing of fluid ingestion. In order to investigate whether this relationship occurred as a result of the osmotic and volumetric effects of the ingested nutrients, analyses were performed on the data obtained from 219 adult humans. They were paid to maintain diaries for 7 days of everything they ingested, the timing and conditions present at the bout, and pre- and postbout serf-ratings of subjective thirst. Carbohydrate and protein intake were found to be the dietary constituents that were most highly related to fluid intake and subjective thirst while sodium and fat were found to be either not at all or only weakly related. Carbohydrate and protein intake were found to be positively related to the amount ingested of total fluid, fluid in excess of digestive requirements, and fluid in the form of "drinks," for the amounts ingested in individual bouts, over the entire day, and over the entire week. In addition, carbohydrate and protein intakes were found to be positively related to the reduction in the subjective state of thirst, while negatively related to the level of thirst self-reported at the end of the bout. The results indicate that fluid intake and subjective thirst are influenced by the repleting characteristics of ingested nutrients and not by their depleting effects, suggesting that fluid intake occurs in response to and as an adjunct of food intake, not fluid homeostasis. Eating
Drinking
Macronutrients
Sodium
Meal pattern
THE regulation of water balance is an essential physiological requirement. It occurs via the balance of intake with utilization and excretion. The intake of fluids is controlled by body fluid homeostasis, in particular the defense of the intracellular (18,22) and extracellular (19,31) fluid compartments, and by nonhomeostatic stimuli such as palatability factors (28), scheduling factors (17), social factors (10), solid food intake (2, 7, 9, 16, 23), anticipation of future deficits (20,26), and ecological conditions (3,29). Fluid intake can be precipitated by any of these homeostatic or nonhomeostatic mechanisms. However, it is not known which of these influences are important to the everyday operation of the system and which operate only rarely or in specialized circumstances. Recent evidence has suggested that, under ad lib conditions, homeostatic mechanisms are not important influences on fluid intake (27). Rather, the amount and timing of fluid intake is primarily determined by the amount and timing of food intake in both rats (9) and humans (7). This suggests that fluid intake, under ad lib conditions, is not regulated, but occurs as an adjunct to eating. Sufficient quantities of fluids are ingested in response to nonregulatory stimuli that a further modification of intake based upon regulatory signals is unnecessary. Body fluid balance is then ac-
Carbohydrate
Fat
Protein
complished by nonhomeostatic drinking of more fluid than is needed and excretion of the excess by the kidney which actually performs the regulation (11). However, this is not the only interpretation of these data. It is possible that fluid ingested along with food is a form of anticipatory drinking (20) in response to the expected fluid depleting/repleting impact of the digestion and metabolism of the ingested food. If this were true, then it would be expected that the amount of fluid ingested would be related to the depleting/repleting characteristics of the coingested nutrients. In particular, sodium ingestion, as a consequence of its marked influence on effective plasma osmolality (35), would be expected to have a large stimulatory effect on fluid ingestion. Hence, the amount of fluid spontaneously ingested along with a meal would be expected to be strongly related to the sodium content of ingested foods. Protein intake likewise would be expected to have a relatively large influence on fluid intake. A diet rich in protein produces a relatively large turnover of water and a consequent increase in fluid intake (25). Carbohydrate would be expected to be relatively less influential since it has only a small osmotic impact and its metabolism results in the production of water rather than its loss (34). Finally, fat intake would be expected to have the least ef-
lRequests for reprints should be addressed to John M. de Castro, Department of Psychology, Georgia State University, University Plaza, Atlanta, GA 30303-3083.
513
514
~1~ C A S T R ( )
fect on fluid intake since it does not have osmotic effects and its metabolism creates water. In order to assess whether the depleting/repleting properties of ingested foods are the determining factor in the control of fluid ingestion, the amounts of each of the macronutrients and sodium ingested in bouts, during each day and over a week, were correlated with univariate and multivariate techniques with the amount of fluids ingested, with the amount of time until the next bout containing fluid, and with the subjects subjective state of thirst at the end of the bout and the change over the bout. It would be expected that if meal-associated fluid intake was an anticipatory response to the expected impact of the ingested nutrients on body fluid balance then the amounts of sodium and protein in the bout should be strongly positively associated with the amounts of fluids ingested, negatively with the time until the next ingestion of fluid, and positively with postbout self-reported thirst, while the amounts of carbohydrate and fat should have much weaker associations. The data needed to evaluate these predictions are available in the data base collected during prior research projects (5-8, 10, 12-15) that is routinely added to with new data from ongoing research projects. These data were collected by asking adult humans to maintain a diary for seven consecutive days of everything they either ate or drank and the time of occurrence. One of the main obstacles to studying spontaneous fluid intake in humans is separating food from fluid intake, since many foods are liquids (e.g., soups), many liquids contain significant amounts of food energy (e.g., milk shake), and even solid foods contain varying amounts of water. In the present study, no attempt was made to draw absolute distinctions but rather three different ways of classifying fluid intakes were employed. In the first analysis, no attempt was made to distinguish food from fluid intake, intake was classified according to its water content and data analyses were performed on simple total water intakes. The second analysis characterized the water intake in terms of its surplus over that required for digestion of the food and analyses were performed on these data. In the last analyses, arbitrary distinctions were made between food and fluid intakes and only the data for those events classified as "drinks" were analyzed. METHOD
The details of the methods used have been published elsewhere (5-7, 14, 15). Therefore, they will only be briefly summarized here.
Subjects Data were collected from 80 male and 139 female adult humans who were recruited from a newspaper ad and by word of mouth and were paid $30 to participate. They also received a detailed nutritional analysis based on their food intake for the 7-day reporting period. The subjects averaged 32.6 years (range 1869), 64.8 kg (range 44.1-102.3) and 1.66 m (ran~ge 1.43-1.90) and had an average body mass index of 23.4 k g / m ' ( r a n g e 18.138.1), Informed consent was obtained from all subjects who were told that the nutrient intakes of humans were being studied. In order to participate the subjects could not be actively dieting, pregnant or lactating, on chronic medication, or alcoholic.
Procedure The subjects were given a small (8 × 18 cm) pocket-sized diary and were instructed to record in as detailed a manner as possible every item that they either ate or drank, the time they ate it,
the amount they ate, and how the food was prepared. One hundred thirty of the subjects were also instructed to record their thirst and hunger at the beginning and again at the end of each bout on seven-point, full-thirsty and full-hungry scales. The subjects recorded for a day and were contacted by the experimenter to review the information, correct any problems, and answer any questions. They then recorded their intake in the diaries for seven consecutive days. After receiving the diaries, the experimenter reviewed them and contacted the subjects to clarity any ambiguities or missing data in the diary records. The subjects were later contacted by phone if any questions arose about their entries in the diaries. Before recording their intake the subjects were asked to provide the names and phone numbers of two individuals who would probably be eating with them sometime during the recording period. After the completed diaries were submitted, each of these individuals were contacted and asked to verify the subjects reported intake. Although some difficulty was encountered in remembering exactly what the subject ate, in no case was the subject's diary report contradicted in either the nature or the amount of the food reported.
Data Analysis The foods reported in the diaries were assigned codes from a computer file of over 3500 food items by an experienced registered dietitian. The coder was unaware of the experimental hypotheses and did not interact directly with the subjects. Bouts were then identified and the compositions of the individual items composing the bout were summed. In order for a reported intake to be classified as an individual bout it had to contain at least 50 ml of liquid, or more stringently 100 or 200 ml. tt also had to be separated in time from the preceding and following ingestive behaviors by at least 15 minutes. More stringent definitions of 45 and 90 minutes were also employed. In order to cover a range of definitions from lenient to strict, five different definitions of a bout were used combining these minimum criteria, 15 rain/50 ml, 45 min/50 ml, 45 min/100 ml, 45 min/200 ml, and 90 min/50 ml. Total daily intakes and the individual bouts were characterized by their contents of total calories, carbohydrate, fat, protein, sodium, total water, excess water, and "drink water" and by their pre- and postbout self-ratings of thirst. Total water was calculated as the amount of water ingested regardless of its source. Water obtained from food and/or drink was simply summed together to yield the total amount of fluid in the bout. Excess water was calculated as the total amount of water ingested above that required for digestion. The amount required for digestion was estimated as equal to the grams of solid ingested times 1.1 [see (32,33) for a discussion of the rationale for the choice of the constant 1.1]. " D r i n k " water was calculated as the amount of liquid contained in events arbitrarily classified as "drinks." In order to qualify as a " d r i n k , " the item had to be ingested in liquid form (e.g., Jello was considered a solid and was not classified as a drink), and had to not be normally considered a food (e.g., soup or instant breakfasts were classified as eating and were not considered "drinks"). For each subject, the amounts ingested of carbohydrate, fat, protein and sodium were correlated using Pearson product moment correlations with the amounts of total, excess, or "drink" water ingested. These same variables were used as independent variables in a multiple linear regression prediction of the amounts of total, excess, or "drink" water ingested. Also, for each subject the amounts ingested of carbohydrate, fat, protein, sodium and total water were correlated with the postbout thirst self-ratings and the pre- to postbout change in the thirst self-ratings. These same variables were used as independent variables in a multiple linear regression prediction of the postbout and pre- to
NUTRIENTS AND FLUID INTAKE
515
TABLE 1 MEANS AND STANDARDERRORSOF THE COMPOSITIONSOF THE DAILY INTAKESAND THE BOUTINTAKESFOR THE 45-min50-ml BOUTDEFINITION
.7
.B
Carbohydrate (g) Fat (g) Protein (g) Sodium (mg) Total water (ml) Excess water (ml) "Drink" water (ml) Postbout thirst Thirst change
Dally Intake
SEM
Bout Intake
SEM
210.4 76.2 75.6 2712.9 1877.3 1513.6 1349.3
5.2 1.8 1.7 64.2 50.5 47.7 46.1
55.5 20.2 20.6 739.4 503.6 397.7 355.9 2.7 - 1.9
1.5 0.6 0.5 22.3 12.0 11.8 11.7 0.1 0.1
~luid
* 0,
.5
•
**
.4
Variable
Total
*
.3 .2 .1 O ,
.7
*
i" u
.3
.!
postbout change in thirst self-ratings. These analyses were performed for the contents of individual bouts and for the overall contents of the ingesta both over the entire day and over the entire 7-day recording period. Group means and standard errors were then calculated using the bout characteristics, beta coefficients from the multiple regressions, and univariate and multivariate correlations that had been calculated for each subject individually. Since correlation coefficients are not normally distributed they were fast transformed to z scores that are normally distributed prior to performing t-tests (4). The mean correlations and coefficients were then compared to 0 with a t-test. Correlations and beta coefficients were compared with correlated t-tests. ~S~TS Analyses were performed on bouts identified by five different definitions of a bout (see Data Analysis section). There were no significant qualitative differences in the results obtained with different definitions. Although data are presented from all definitions, the descriptive and inferential statistics reported in the text are for the minimum 50 mi, 45 min definition, which is presented as representative. The mean overall and bout intakes for all the subjects are presented in Table 1. The subjects ingested on average 1830---45.2 kcal per day in 3.8--.0.08 bouts consisting of 57.6% carbohydrate, 21.0% fat, and 21.4% protein. The uuivariate correlations between the dietary constituents and the fluid intake variables are presented in Fig. 1. The correlations calculated between the nutrient and the fluid intakes over the entire day are represented by the solid (filled) bars in Fig. 1. Although the magnitude of the correlations varies according to whether total fluid (top), excess fluid (middle) or "drink" fluid (bottom) are used for the correlations, the same pattern emerges. Carbohydrate has the strongest relationship with fluid intake compared to fat [t(218)=6.86, 5.75, 6.36, p<0.05], compared to protein [t(218)= 3.49, 1.68 (n.s.), 4.94, p<0.05], and compared to sodium [t(218)= 5.53, 2.86, 6.36, p<0.05] for total, excess, and "drink" fluid, respectively. On the other hand fat has the weakest relationship with fluid intake compared to protein [t(218) = 3.86, 4.72, 1.38 (n.s.), p
g I
o ,
"7.1 ,8
Drink Eloid only
N'
. * (t
m
Daily I n t a k e
Prediction
Weekly I n t o ~ o P r e d l c t l o n
.5 .4 .3 .2 .1 0
Carbohydrat~
Foe
Proteir~
Sodium
HG. 1. Mean_ SEM correlations between coingested carbohydrate (left 7 bars), fat (middle left 7 bars), protein (middle fight 7 bars), and sodium (right 7 bars) and the amounts of total fluid (top), fluid in excess of digestive requirements (middle), and fluid ingested in the form of "drinks" (bottom) in the bouts (5 hatched bars), in the total intakes over the day (solid bar) and over the week (cross-hatched bar). The first bar of each set of five hatched bars representing the intake in bouts represents the bout definition of minimum 15 minute IMI and 50 ml size; the second, 45 rain/50 ml; the third, 45 rain/100 ml; the fourth, 45 min/200 ml; and the fifth, 90 min/50 ml. *Indicates that the mean is significantly (p<0.05) different from zero as assessed with a t-test.
The correlations calculated between the nutrient and the fluid intakes for individual bouts are represented by the hatched bars in Fig. 1. The pattern is very similar to that found for intakes over the entire day. Carbohydrate again has the strongest relationship with fluid intake compared to fat [t(218)=5.80, 5.78, 8.51, p
516
~÷ CASTRO
Total ,5
CorroJatllon$
Gluid
T * *
$
,
,
.3
.1
o
Z
gz -.2
Fluid • u
* * ,
,
o
.3
Carbohydrato
Fat
Protllm
;odi um
o"
u
.i
-.i
Multiple
Regreosiono
÷
g
-.3
1 Orink .5 .4
Pluid onl X [] MeQl Int~H~ Pr~dictlon I Doily Intake Prediction [] Weekly I n t o ~ Prediction
.3 * *
* , ,
.!
-.| -.2
Carbohydrate
Carbohydrat~
Fat
Protein
Fat
Pro~eim
Sodzum
5od~ um
FIG. 2. Mean_ SEM beta coefficients from the multiple linear regressions with coingested carbohydrate (left 7 bars), fat (middle left 7 bars), protein (middle right 7 bars), and sodium (right 7 bars) as independent variables predicting the dependent variables of the amounts of total fluid (top), fluid in excess of digestive requirements (middle), and fluid ingested in the form of "drinks" (bottom) in the bouts (5 hatched bars), in the total intakes over the day (solid bar) and over the week (cross-hatched bar). The first bar of each set of five hatched bars representing the intake in bouts represents the bout definition of minimum 15 minute 1MI and 50 ml size; the second, 45 min/50 mi; the third, 45 min/100 mi; the fourth, 45 rain/200 ml; and the fifth, 90 min/50 ml. *Indicates that the mean is significantly (p<0.05) different from zero as assessed with a t-test. The multiple regressions calculated between the nutrient and the fluid intakes over the entire day are represented by the solid (filled) bars in Fig. 2. Although the magnitude of the beta coefficients varies according to whether total fluid (top), excess fluid (middle) or " d r i n k " fluid (bottom) is used for the dependent variable, once again, as with the correlations, the same pattern emerges. Carbohydrate has the strongest relationship w i t h fluid intake compared to fat [t(218) = 3.14, 2.88, 2.84, p<0.05], compared to protein [t(218)=2.16, 1.28 (n.s.), 3.16, p<0.05], and compared to sodium [t(218)=2.51, 1.26 (n.s.), 2.76, p<0.05] for total, excess, and "&'ink" fluid, respectively. On the other hand fat has the weakest relationship with fluid intake compared to protein [t(218)=2.68, 2.75, 1.36 (n.s.), p<0.05] and compared to sodium [t(218)=2.54, 2.96, 1.70 (n.s.), p<0.05] for total, excess, and " d r i n k " fluid, respectively. In no case do protein and sodium significantly differ, The beta coefficients calculated over the entire weeks intakes are represented by crosshatched bars in Fig. 2. The pattern for these coefficients is very similar to that for the daily intakes except that the carbohydrate beta coefficients are considerably smaller while for protein they are larger.
FIG. 3. Mean± SEM correlations (top) and beta coefficients from the multiple linear regressions (bottom) between coingested carbohydrate (left 5 bars), fat (middle left 5 bars), protein (middle right 5 bars), and sodium (right 5 bars) and the amount of time until the next bout containing fluid. The first bar of each set of five represents the bout definition of minimum 15 minute IMI and 50 ml size; the second, 45 min/50 ml; the thirdi 45 min/100 mi; the fourth, 45 min/200 ml; and the fifth, 90 min/50 ml. *Indicates that the mean is significantly (p<0.05) different from zero as assessed with a t-test.
The beta coefficients from the multiple regressions calculated between the nutrient and the fluid intakes for individual bouts are represented by the hatched bars in Fig. 2. The pattern is very similar to that found for intakes over the entire day and for the univariate correlational analyses. Carbohydrate again has the strongest relationship with fluid intake compared to fat [t(218)= 12.71, 12.92, 11.94, p<0.05], compared to protein [t(218)= 1.32 (n.s.), 1.52 (n.s.), 7.12, p<0.05], and compared to sodium [t(218) = 4.52, 3.97, 11.53, p<0.05] for total, excess, and "drink" fluid, respectively. Again fat has the weakest relationship with fluid intake compared to protein [t(218)= 11.96, 12.23. 4.21, p<0.05] and compared to sodium [t(218)=7.54, 8.27. 7.12. p<0.05] for total, excess, and " d r i n k " fluid, respectively. Protein has significantly stronger correlations than sodium [t(218)= 5.47, 5.14, 3.79, p<0.05] for total, excess, and " d r i n k " fluid. respectively. The macronutrients and sodium ingested in the bouts were also used to predict the amount of time that would elapse before the next occasion of fluid ingestion (after bout interval). The univariate correlations (top) and the beta coefficients from the multiple linear regression analyses (bottom) are presented in Fig. 3. The macronutrients and sodium all have small but significant positive correlations with the after bout interval. There were no
NUTRIENTS AND FLUID INTAKE
Poeebout
$elf-Roted
517
T~lr~t
Pre
Correlotto~a 0
-.1
to
Poe¢bout
Chnnge i n $mlF-Ru~md
Thiret
Cerrelatlor~
g
.3
1 ii
7, z E
o~ 0
z z z 7, z
.2
0
=[
°
.
g
I
-.3
Mul t l p l e RegreaeJone
Multlpl@ RRgr@m@Ion@
tJ
o
*
,
o~
I -.2
Ccn-t~ol~ya~ate
Foe
ProgeJn
$odltffe
TOrn]
Wnt~
FIG. 4. Mean +-SEM correlations (top) and beta coefficients from the multiple linear regressions (bottom) between coingested carbohydrate (left 5 bars), fat (middle left 5 bars), protein (middle 5 bars), sodium (middle right 5 bars), and total water (fight 5 bars) and the self-ratings of subjective thirst at the end of the bout. The first bar of each set of five represents the bout definition of minimum 15 minute IMI and 50 ml size; the second, 45 rain/50 ml; the third, 45 rain/100 ml; the fourth, 45 rain/200 ml; and the fifth, 90 rain/50 mi. *Indicates that the mean is significantly (,o<0.05) different from zero as assessed with a t-test. significant differences between the magnitudes of these correlations. Similarly, the beta coefficients from the multiple regressions predicting the after bout interval were all small and positive but only rarely significant. There were also no significant differences between the magnitudes of these beta coefficients. The macronutrients, sodium and total water ingested in the bouts were also used to predict the level of self-rated thirst at the end of the bout. The univariate correlations (top) and the beta coefficients from the multiple linear regression analyses (bottom) are presented in Fig. 4. The nutrients, sodium, and total water all have significant negative correlations with postbout self-rated thirst, the more of each the lower the thirst. Protein had significantly larger correlations than any of the other constituents of the bouts [t(129) = 3.07, 2.10, 1,99, 2.53, p<0.05] for carbohydrate, fat, sodium and total water, respectively. There were no significant differences between the correlations obtained for any of the other constituents. The beta coefficients from the multiple regressions also indicated negative relationships between the constituents and thirst but only in the cases of carbohydrate, protein and total water were the beta coefficients significantly different from zero. These results indicate the more carbohydrate, protein, or total water ingested in a bout the lower the thirst self-rating at the end of the bout. These same factors were used as predictors of the change in
-. 1
FIG. 5. Mean__.SEM correlations (top) and beta coefficients from the multiple linear regressions (bottom) between coingested carbohydrate (left 5 bars), fat (middle left 5 bars), protein (middle 5 bars), sodium (middle fight 5 bars), and total water (fight 5 bars) and the change in the self-ratings of subjective thirst from prior to the bout to the end of the bout. The f'wst bar of each set of five represents the bout definition of minimum 15 minute IMI and 50 ml size; the second, 45 rain/50 ml; the third, 45 rain/ 100 ml; the fourth, 45 min/200 ml; and the fifth, 90 rain/50 mi. *Indicates that the mean is significantly (p
self-rated thirst from the start of the bout to the end. The unlvariate correlations (top) and the beta coefficients from the multiple linear regression analyses (bottom) are presented in Fig. 5. The nutrients, sodium, and total water all have significant positive correlations with the change in thirst, the more of each the greater the reduction in thirst occurring during the bout. The only significant difference between these correlations was that the correlation with fat intake was significantly lower than that with either protein, sodium, or total water [t(129)=2.61, 2.21, 2.14, p<0.05], respectively. The multiple regression analysis produced beta coefficients which were positive and significantly different from zero for carbohydrate, protein, and total water, indicating that the more of these constituents present the greater the reduction in thirst occuring over the course of the bout. The beta coefficients for fat significantly differed from carbohydrate, protein, and total water [t(129)= 2.79, 2.26, 3.81, p<0.05], respectively. DISCUSSION The present results indicate that the fluid ingested during bouts or over the course of a day or a week is not primarily related to the depleting/repleting characteristics of the coingested nutrients. The amount of time elapsing until the next bout, also, does not
518
appear to be influenced by the nutrients' impact on fluid balance. Additionally, the subjective state of thirst, although greatly affected by the ingestion of nutrients, does not appear to be related, either, to the nutrients' impact on body fluids. These results, then, suggest that fluid intake is not anticipatory of the fluid depleting/ repleting impact of the digestion and metabolism of ingested food. They further suggest that under ad lib conditions in the natural human environment the intake of fluids and the subjective state of thirst would not appear to be primarily determined by fluid homeostasis. The expected large impact of sodium ingestion was not present. Although sodium intake correlates with fluid ingestion, this correlation would appear to result from the covariation of sodium intake with the overall meal size. When sodium intake was used in a multiple regression along with the three macronutrients its relationship with either total fluid or excess fluid ingested became small and with "drink" fluid vanished entirely. This was true regardless of whether the intakes occurring in bouts, over the entire day, or even over the entire week were used in the calculations. Additionally, it was expected that if the influence of ingested sodium on the effective plasma osmolality was an influence on the regulation of fluid intake then a negative relationship would be expected between sodium intake and the after bout interval. It was expected that the more sodium ingested in the bout the shorter the interval until fluid was again ingested. In fact, the opposite was found with the amount of sodium in the bout positively associated with the duration of the after bout interval. Sodium ingestion also did not have the expected effect on the subjective state of thirst. It would be expected that sodium would be positively correlated with thirst and negatively related to the satisfaction of thirst, i.e., the change scores. In fact, the opposite was found with the amount of sodium ingested negatively correlated with postbout thirst and positively correlated with the change in thirst over the bout. Indeed, even when the influence of the amount of water ingested in the bout was removed in the multiple regression, sodium intake failed to show the expected influence on the subjective state. Thus neither the amount of fluid ingested, its timing, nor the subjective state of thirst or its relief would appear to be strongly influenced by ingested sodium. Thus it does not appear that thirst and drinking are anticipatory to the postingestive effects of sodium on the effective plasma osmolality (35) whether looked at over meal, daily, or weekly time spans. These results, then, call into question the importance of plasma osmolality as a major determinant of either ad lib fluid intake or the subjective state of thirst. Protein intake, since it produces a relatively large turnover of water, was expected to be strongly related to fluid intake as has been observed in manipulative experimental studies (25). To a moderate extent, the expected relationships with fluid ingestion were found but not with the timing of intake or the subjective state of thirst. Protein intake was found to be positively related to total and excess fluid intake ingested in bouts, over the day, and over the week with both univariate and multivariate analyses, but only weakly related with fluids ingested in the form of "drinks." However, in many ways protein ingestion was not found to have the expected relationships. It was expected that since protein would have a fluid-depleting effect, then the more protein ingested in the bout the shorter the interval until fluid was again ingested. This was not the case. Indeed, protein ingestion was positively related to the duration of the postbout interval. Additionally, protein ingestion would be expected to be associated with higher postbout thirst and a smaller change in subjective thirst over the bout. In fact the opposite was found. Protein intake not only had a negative association with postbout thirst but, in fact, had the strongest negative association of all of the constituents looked at. Thus, once again, the present results do not
~k CASTRO
support the notion that anticipation of changes in body fluid homeostasis is the determining factor in the regulation of fluid mtake and subjective thirst. Carbohydrate intake, since it has only a small osmotic impact (34), was not expected to be strongly related to fluid intake or to the subjective state of thirst. Contrary to this prediction, carbohydrate, of all the measured constituents, was probably the most strongly related to fluid intake and was among the strongest in association with the postbout interval and the subjective state. The only constituent that did have the expected relationships with fluid intake and thirst was fat. Based upon the fact that fat ingestion does not have osmotic effects, it would not be expected to b¢ related to either fluid intake or thirst. Indeed, fat intake was found to have little discernable association with either the amount of fluid ingested, the timing of fluid intake, or the subjective state of thirst at the end of the bout or its change over the bout. The present results, then, clearly do not support an explanation of fluid intake based upon the anticipation of changes in body fluid balance resulting from ingested nutrients. Intake and subjective state were not related to the osmotic or volumetric properties of the coingesta. What, then, determines fluid intake and thirst? Clearly, the foods normally ingested by humans in their normal everyday existences are highly hydrated and in combination with "drinks" taken along with the foods produce a marked surfeit of fluids over that needed for digestion. This is evidenced by the fact that over 1.5 liters of water per day, over 80% of the total daily amount of fluid, was ingested in excess of digestive requirements (Table 1). Since, the nutrients are so highly hydrated, it would seem reasonable that their ingestion would be positively related to the satisfaction of subjective thirst. Additionally, this high degree of hydration would mean that any depleting effects of the nutrients would be more than compensated for. The results make more sense when conceptualized in reference to the fluid repletion effects of the bout rather than depletion effects. Since fat and sodium are normally not hydrated, the fact that they were not related to thirst and fluid intake makes sense, given they are not contributing any fluid to the system. Carbohydrate and protein in the human diet, on the other hand, are heavily hydrated and thus contribute to repletion rather than depletion. Thus the positive relationships of these two nutrients with fluid intake and thirst reduction are readily understandable. In prior papers, it was suggested that fluid intake was mainly governed by feeding in the ad lib pattern of intake of both rats and humans (7, 9, 10). Fluid was seen as ingested well above requirements and the job of maintaining body fluid balance was left to the kidney. The present results support this interpretation. In addition, the present results suggest that fluid intake may be primarily governed by palatability factors, with both the palatability of the "drinks" and the solid foods involved. The clear and strong relationship of carbohydrate with the amount of fluid ingested suggests that sweetness may be an important determinant. Since ingested "drinks" frequently contain carbohydrate and since ingested fluids increase the palatability of ingested foods, palatability effects alone could account for the covariation of food and fluid intake in the natural ad lib pattern of ingestive behaviors. Self-reports of food intake have traditionally been thought to be inaccurate. In the case of diet recall procedures there probably is considerable inaccuracy (24). However, the diet diary technique where subjects record at the same time that they eat has been demonstrated to be both reliable and valid (1, 21, 24, 30). Additionally, in the present study the subjects' reward for accurate record keeping was a detailed analysis of their reported di, ets. The subjects expressed great interest in receiving such an analysis and were aware that it would only be as accurate as their records. Also, the subject's diary entries were verified by two people who ate with the subject. Thus there is every, reason to
NUTRIENTS AND FLUID INTAKE
519
believe that the self-reports acquired in the present study are accurate. Even if the technique produces many errors, there is no reason to suspect any systematic relationship between recording errors and the primary variables of interest in the study. Unsystematic error should cover up significant relationships not produce them. The fact that clear, replicable, and highly significant results, as are reported in the present study, can be discerned with a somewhat insensitive and error prone technique speaks to the power and importance of the effects reported. The present results clearly indicate that under ad lib conditions in humans fluid intake and subjective thirst are not related to the depleting characteristics of the ingested nutrients. Rather, fluid intake and thirst are related to their repleting characteristics. This again supports a model that ad lib fluid intake occurs in response to and as an adjunct of food intake, that as a result fluid is ingested in excess of requirements, and that the kidney is primarily responsible for body fluid homeostasis. This interpretation does not necessarily apply to other ecological conditions. The data for
the present study were obtained from primarily middle class Americans living in a m o d e m urban environment where food and fluids are readily and abundantly available. In situations where fluids are not available ad lib or where obtaining water might be difficult or dangerous, tighter regulation of fluid intake independent of food intake may well occur (3,29).
ACKNOWLEDGEMENTS The author would like to express his appreciation and acknowledge the substantial contributions of Ms. Dixie K. Elmore, Marie Brewer, and Ms. Margaret Pedersen without whose assistance the work could not have been performed, and to Dr. Terry Thrasher and Dr. Gary Robertson for suggesting the analysis. This research was supported in part by Grant R01-DK39881-01A2 from the National Institute of Diabetes and Digestive and Kidney Diseases, from a grant from the Georgia State University Research Grant Program and from a Biological Research Support Grant whose support is gratefully acknowledged.
REFERENCES 1. Adleson, S. F. Some problems in collecting dietary data from individuals. J. Am. Diet. Assoc. 36:453-461; 1960. 2. Bigelow, J. E.; Houpt, T. R. Feeding and drinking patterns in young pigs. Physiol. Behav. 43:99-109; 1988. 3. Collier, G. The dialog between the house economist and the resident physiologist. Nutr. Behav. 3:9-26; 1986. 4. de Castro, J. M. Meal pattern correlations: Facts and artifacts. Physiol. Behav. 15:13-15; 1975. 5. de Castro, J. M. Macronutrient relationships with meal patterns and mood in the spontaneous feeding behavior of humans. Physiol. Behav. 39:561-569; 1987. 6. de Castro, J. M. Circadian rhythms of the spontaneous meal patterns, macronutrient intake, and mood of humans. Physiol. Behav. 40:437466; 1987. 7. de Castro, J. M. A microregulatory analysis of spontaneous fluid intake by humans: Evidence that the amount of liquid ingested and its timing is mainly governed by feeding. Physiol. Behav. 43:705-714; 1988. 8. de Castro, J. M. Physiological, environmental, and subjective determinants of food intake in humans: A meal pattern analysis. Physiol. Behav. 44:651-659; 1988. 9. de Castro, J. M. The interaction of fluid and food intake in the spontaneous feeding and drinking patterns of rats. Physiol. Behav. 45: 861-870; 1989. 10. de Castro, J. M. Social, circadian, nutritional, and subjective correlates of the spontaneous pattern of moderate alcohol intake of normal humans. Pharmacol. Biochem. Behav. 35:923-931; 1990. 11. de Castro, J, M. Bout pattern analysis of ad lib fluid intake. In: Ramsay, D. J.; Booth, D. A., eds. Thirst: Physiological and psychological aspects. Bedim Springer-Verlag; 1991:in press. 12. de Castro, J. M.; Brewer, M.; Elmore, D. K.; Orozco, S. Social facilitation of the spontaneous meal size of humans is independent of time, place, alcohol, or snacks. Appetite 15:89-101; 1990. 13. de Castro, J. M.; de Castro, E. S. Spontaneous meal patterns of humans: influence of the presence of other people. Am. J. Clin. Nutr. 50:237-247; 1989. 14. de Castro, J. M.; Elmore, D. K. Subjective hunger relationships with meal patterns in the spontaneous feeding behavior of humans: Evidence for a causal connection. Physiol. Behav. 43:159-165; 1988. 15. de Castro, J. M.; McCormick, J.; Pedersen, M.; Kreitzman, S. N. Spontaneous human meal patterns are related to preprandial factors regardless of natural environmental constraints. Physiol. Behav. 38: 25-29; 1986. 16. Engell, D. Interdependency of food and water intake in humans. Appetite 10:133-141; 1988. 17. Falk, J. L. Production of polydipsia in normal rats by an intermittent
food schedule. Science 133:195-196; 1966. 18. Fitzsimons, J. T. Drinking by nephrectomized rats injected with various substances. J. Physiol. (Lond.) 155:563-579; 1961. 19. Fitzsimons, J. T. Drinking of rats depleted of body fluid without increase in osmotic pressure, J. Physiol. (Lond.) 159:297-309; 1961. 20. Fitzsimons, J. T.; LeMagnen, J. Eating as a regulatory control of drinking. J. Comp. Physiol. Psychol. 67:273-283; 1969. 21. Gersovitz, M.; Madden, J. P.; Smicikalas-Wright, H. Validity of the 24-hour dietary recall and seven-day record for group comparisons. J. Am. Diet. Assoc. 73:48-55; 1978. 22. Gilman, A. The relation between blood osmotic pressure, fluid distribution and voluntary water intake. Am. J. Physiol. 120:323-328; 1937. 23. Kissileff, H. R. Food associated drinking in the rat. J. Comp. Physiol. Psychol. 67:284-300; 1969. 24. Krantzler, N. J.; Mullen, B. J.; Schultz, H. G.; Grivetti, L. E.; Holden, C. A.; Meiselman, H. L. The validity of telephoned diet recalls and records for assessment of individual food intake. Am. J. Clin. Nutr. 36:1234-1242; 1982. 25. Le Magnen, J.; Tallon, S. Les EMterminants quantitatifs de la prise hydrique clans ses relations avec la prise d'aliments chez le rat. C R Soc. Biol. (Paris) 161:1243-1246; 1967. 26. Lucas, G. A.; Timberlake, W.; Gawley, D. J. Learning and meaiassociated drinking: meal-related deficits produce adjustments in postprandial drinking. Physiol. Behav. 46:361-367; 1989. 27. Phillips, P. A.; Rolls, B. J.; Ledingham, J. G. G.; Morton, J. J. Body fluid changes, thirst and drinking in man during free access to water. Physiol. Behav. 33:357-363; 1984. 28. Rolls, B. J.; Wood, R. J.; Stevens, R. M. Effects of palatability on body fluid homeostasis. Physiol. Behav. 20:15-19; 1978. 29. Rowland, N. Regulatory drinking: do physiological substrates have an ecological niche? Biobehav. Rev. 1:261-272; 1977. 30. St. Jeor, S. T.; Guthrie, H. A.; Jones, M. B. Variability of nutrient intake in a 28 day period. J. Am. Diet. Assoc. 83:155-162; 1983. 31. Striker, E. M. Some physiological and motivational properties of the hypovolemic stimulus for thirst. Physiol. Behav. 3:379-385; 1968. 32. Toates, F. M. A physiological control theory of the hunger-thirst interaction. In: Booth, D. A., ed. Hunger models. New York: Academic Press; 1978:347-374. 33. Toates, F. M. Computer simulation and the homeostatic control of behaviour. In: McFarland, D. J., ed. Motivational control systems analysis. New York: Academic Press; 1974:407--426. 34. Voices, T. Water homeostasis. Annu. Rev. Nutr. 7:383---406; 1987. 35. Zerbe, R. L.; Robertson, G. L. Osmoregulation of thirst and vasopressin secretion in human subjects: effects of various solutes. Am. J. Physiol. 244:E607-E614; 1983.
0031-9384/91 $3.00 + .00
The Relationship of Spontaneous Macronutrient and Sodium Intake With Fluid Ingestion and Thirst in Humans J O H N M. DE C A S T R O 1
Department of Psychology and the Behavior and Neurobiology Program Georgia State University, Atlanta, GA 30303-3083 Received 2 July 1990
DE CASTRO, J. M. The relationship of spontaneous macronutrient and sodium intake with fluid ingestion and thirst in humans. PHYSIOL BEHAV 49(3) 513-519, 1991.--The amount of solid food eaten by humans in spontaneously ingested bouts is the most important determinant of the amount and timing of fluid ingestion. In order to investigate whether this relationship occurred as a result of the osmotic and volumetric effects of the ingested nutrients, analyses were performed on the data obtained from 219 adult humans. They were paid to maintain diaries for 7 days of everything they ingested, the timing and conditions present at the bout, and pre- and postbout serf-ratings of subjective thirst. Carbohydrate and protein intake were found to be the dietary constituents that were most highly related to fluid intake and subjective thirst while sodium and fat were found to be either not at all or only weakly related. Carbohydrate and protein intake were found to be positively related to the amount ingested of total fluid, fluid in excess of digestive requirements, and fluid in the form of "drinks," for the amounts ingested in individual bouts, over the entire day, and over the entire week. In addition, carbohydrate and protein intakes were found to be positively related to the reduction in the subjective state of thirst, while negatively related to the level of thirst self-reported at the end of the bout. The results indicate that fluid intake and subjective thirst are influenced by the repleting characteristics of ingested nutrients and not by their depleting effects, suggesting that fluid intake occurs in response to and as an adjunct of food intake, not fluid homeostasis. Eating
Drinking
Macronutrients
Sodium
Meal pattern
THE regulation of water balance is an essential physiological requirement. It occurs via the balance of intake with utilization and excretion. The intake of fluids is controlled by body fluid homeostasis, in particular the defense of the intracellular (18,22) and extracellular (19,31) fluid compartments, and by nonhomeostatic stimuli such as palatability factors (28), scheduling factors (17), social factors (10), solid food intake (2, 7, 9, 16, 23), anticipation of future deficits (20,26), and ecological conditions (3,29). Fluid intake can be precipitated by any of these homeostatic or nonhomeostatic mechanisms. However, it is not known which of these influences are important to the everyday operation of the system and which operate only rarely or in specialized circumstances. Recent evidence has suggested that, under ad lib conditions, homeostatic mechanisms are not important influences on fluid intake (27). Rather, the amount and timing of fluid intake is primarily determined by the amount and timing of food intake in both rats (9) and humans (7). This suggests that fluid intake, under ad lib conditions, is not regulated, but occurs as an adjunct to eating. Sufficient quantities of fluids are ingested in response to nonregulatory stimuli that a further modification of intake based upon regulatory signals is unnecessary. Body fluid balance is then ac-
Carbohydrate
Fat
Protein
complished by nonhomeostatic drinking of more fluid than is needed and excretion of the excess by the kidney which actually performs the regulation (11). However, this is not the only interpretation of these data. It is possible that fluid ingested along with food is a form of anticipatory drinking (20) in response to the expected fluid depleting/repleting impact of the digestion and metabolism of the ingested food. If this were true, then it would be expected that the amount of fluid ingested would be related to the depleting/repleting characteristics of the coingested nutrients. In particular, sodium ingestion, as a consequence of its marked influence on effective plasma osmolality (35), would be expected to have a large stimulatory effect on fluid ingestion. Hence, the amount of fluid spontaneously ingested along with a meal would be expected to be strongly related to the sodium content of ingested foods. Protein intake likewise would be expected to have a relatively large influence on fluid intake. A diet rich in protein produces a relatively large turnover of water and a consequent increase in fluid intake (25). Carbohydrate would be expected to be relatively less influential since it has only a small osmotic impact and its metabolism results in the production of water rather than its loss (34). Finally, fat intake would be expected to have the least ef-
lRequests for reprints should be addressed to John M. de Castro, Department of Psychology, Georgia State University, University Plaza, Atlanta, GA 30303-3083.
513
514
~1~ C A S T R ( )
fect on fluid intake since it does not have osmotic effects and its metabolism creates water. In order to assess whether the depleting/repleting properties of ingested foods are the determining factor in the control of fluid ingestion, the amounts of each of the macronutrients and sodium ingested in bouts, during each day and over a week, were correlated with univariate and multivariate techniques with the amount of fluids ingested, with the amount of time until the next bout containing fluid, and with the subjects subjective state of thirst at the end of the bout and the change over the bout. It would be expected that if meal-associated fluid intake was an anticipatory response to the expected impact of the ingested nutrients on body fluid balance then the amounts of sodium and protein in the bout should be strongly positively associated with the amounts of fluids ingested, negatively with the time until the next ingestion of fluid, and positively with postbout self-reported thirst, while the amounts of carbohydrate and fat should have much weaker associations. The data needed to evaluate these predictions are available in the data base collected during prior research projects (5-8, 10, 12-15) that is routinely added to with new data from ongoing research projects. These data were collected by asking adult humans to maintain a diary for seven consecutive days of everything they either ate or drank and the time of occurrence. One of the main obstacles to studying spontaneous fluid intake in humans is separating food from fluid intake, since many foods are liquids (e.g., soups), many liquids contain significant amounts of food energy (e.g., milk shake), and even solid foods contain varying amounts of water. In the present study, no attempt was made to draw absolute distinctions but rather three different ways of classifying fluid intakes were employed. In the first analysis, no attempt was made to distinguish food from fluid intake, intake was classified according to its water content and data analyses were performed on simple total water intakes. The second analysis characterized the water intake in terms of its surplus over that required for digestion of the food and analyses were performed on these data. In the last analyses, arbitrary distinctions were made between food and fluid intakes and only the data for those events classified as "drinks" were analyzed. METHOD
The details of the methods used have been published elsewhere (5-7, 14, 15). Therefore, they will only be briefly summarized here.
Subjects Data were collected from 80 male and 139 female adult humans who were recruited from a newspaper ad and by word of mouth and were paid $30 to participate. They also received a detailed nutritional analysis based on their food intake for the 7-day reporting period. The subjects averaged 32.6 years (range 1869), 64.8 kg (range 44.1-102.3) and 1.66 m (ran~ge 1.43-1.90) and had an average body mass index of 23.4 k g / m ' ( r a n g e 18.138.1), Informed consent was obtained from all subjects who were told that the nutrient intakes of humans were being studied. In order to participate the subjects could not be actively dieting, pregnant or lactating, on chronic medication, or alcoholic.
Procedure The subjects were given a small (8 × 18 cm) pocket-sized diary and were instructed to record in as detailed a manner as possible every item that they either ate or drank, the time they ate it,
the amount they ate, and how the food was prepared. One hundred thirty of the subjects were also instructed to record their thirst and hunger at the beginning and again at the end of each bout on seven-point, full-thirsty and full-hungry scales. The subjects recorded for a day and were contacted by the experimenter to review the information, correct any problems, and answer any questions. They then recorded their intake in the diaries for seven consecutive days. After receiving the diaries, the experimenter reviewed them and contacted the subjects to clarity any ambiguities or missing data in the diary records. The subjects were later contacted by phone if any questions arose about their entries in the diaries. Before recording their intake the subjects were asked to provide the names and phone numbers of two individuals who would probably be eating with them sometime during the recording period. After the completed diaries were submitted, each of these individuals were contacted and asked to verify the subjects reported intake. Although some difficulty was encountered in remembering exactly what the subject ate, in no case was the subject's diary report contradicted in either the nature or the amount of the food reported.
Data Analysis The foods reported in the diaries were assigned codes from a computer file of over 3500 food items by an experienced registered dietitian. The coder was unaware of the experimental hypotheses and did not interact directly with the subjects. Bouts were then identified and the compositions of the individual items composing the bout were summed. In order for a reported intake to be classified as an individual bout it had to contain at least 50 ml of liquid, or more stringently 100 or 200 ml. tt also had to be separated in time from the preceding and following ingestive behaviors by at least 15 minutes. More stringent definitions of 45 and 90 minutes were also employed. In order to cover a range of definitions from lenient to strict, five different definitions of a bout were used combining these minimum criteria, 15 rain/50 ml, 45 min/50 ml, 45 min/100 ml, 45 min/200 ml, and 90 min/50 ml. Total daily intakes and the individual bouts were characterized by their contents of total calories, carbohydrate, fat, protein, sodium, total water, excess water, and "drink water" and by their pre- and postbout self-ratings of thirst. Total water was calculated as the amount of water ingested regardless of its source. Water obtained from food and/or drink was simply summed together to yield the total amount of fluid in the bout. Excess water was calculated as the total amount of water ingested above that required for digestion. The amount required for digestion was estimated as equal to the grams of solid ingested times 1.1 [see (32,33) for a discussion of the rationale for the choice of the constant 1.1]. " D r i n k " water was calculated as the amount of liquid contained in events arbitrarily classified as "drinks." In order to qualify as a " d r i n k , " the item had to be ingested in liquid form (e.g., Jello was considered a solid and was not classified as a drink), and had to not be normally considered a food (e.g., soup or instant breakfasts were classified as eating and were not considered "drinks"). For each subject, the amounts ingested of carbohydrate, fat, protein and sodium were correlated using Pearson product moment correlations with the amounts of total, excess, or "drink" water ingested. These same variables were used as independent variables in a multiple linear regression prediction of the amounts of total, excess, or "drink" water ingested. Also, for each subject the amounts ingested of carbohydrate, fat, protein, sodium and total water were correlated with the postbout thirst self-ratings and the pre- to postbout change in the thirst self-ratings. These same variables were used as independent variables in a multiple linear regression prediction of the postbout and pre- to
NUTRIENTS AND FLUID INTAKE
515
TABLE 1 MEANS AND STANDARDERRORSOF THE COMPOSITIONSOF THE DAILY INTAKESAND THE BOUTINTAKESFOR THE 45-min50-ml BOUTDEFINITION
.7
.B
Carbohydrate (g) Fat (g) Protein (g) Sodium (mg) Total water (ml) Excess water (ml) "Drink" water (ml) Postbout thirst Thirst change
Dally Intake
SEM
Bout Intake
SEM
210.4 76.2 75.6 2712.9 1877.3 1513.6 1349.3
5.2 1.8 1.7 64.2 50.5 47.7 46.1
55.5 20.2 20.6 739.4 503.6 397.7 355.9 2.7 - 1.9
1.5 0.6 0.5 22.3 12.0 11.8 11.7 0.1 0.1
~luid
* 0,
.5
•
**
.4
Variable
Total
*
.3 .2 .1 O ,
.7
*
i" u
.3
.!
postbout change in thirst self-ratings. These analyses were performed for the contents of individual bouts and for the overall contents of the ingesta both over the entire day and over the entire 7-day recording period. Group means and standard errors were then calculated using the bout characteristics, beta coefficients from the multiple regressions, and univariate and multivariate correlations that had been calculated for each subject individually. Since correlation coefficients are not normally distributed they were fast transformed to z scores that are normally distributed prior to performing t-tests (4). The mean correlations and coefficients were then compared to 0 with a t-test. Correlations and beta coefficients were compared with correlated t-tests. ~S~TS Analyses were performed on bouts identified by five different definitions of a bout (see Data Analysis section). There were no significant qualitative differences in the results obtained with different definitions. Although data are presented from all definitions, the descriptive and inferential statistics reported in the text are for the minimum 50 mi, 45 min definition, which is presented as representative. The mean overall and bout intakes for all the subjects are presented in Table 1. The subjects ingested on average 1830---45.2 kcal per day in 3.8--.0.08 bouts consisting of 57.6% carbohydrate, 21.0% fat, and 21.4% protein. The uuivariate correlations between the dietary constituents and the fluid intake variables are presented in Fig. 1. The correlations calculated between the nutrient and the fluid intakes over the entire day are represented by the solid (filled) bars in Fig. 1. Although the magnitude of the correlations varies according to whether total fluid (top), excess fluid (middle) or "drink" fluid (bottom) are used for the correlations, the same pattern emerges. Carbohydrate has the strongest relationship with fluid intake compared to fat [t(218)=6.86, 5.75, 6.36, p<0.05], compared to protein [t(218)= 3.49, 1.68 (n.s.), 4.94, p<0.05], and compared to sodium [t(218)= 5.53, 2.86, 6.36, p<0.05] for total, excess, and "drink" fluid, respectively. On the other hand fat has the weakest relationship with fluid intake compared to protein [t(218) = 3.86, 4.72, 1.38 (n.s.), p
g I
o ,
"7.1 ,8
Drink Eloid only
N'
. * (t
m
Daily I n t a k e
Prediction
Weekly I n t o ~ o P r e d l c t l o n
.5 .4 .3 .2 .1 0
Carbohydrat~
Foe
Proteir~
Sodium
HG. 1. Mean_ SEM correlations between coingested carbohydrate (left 7 bars), fat (middle left 7 bars), protein (middle fight 7 bars), and sodium (right 7 bars) and the amounts of total fluid (top), fluid in excess of digestive requirements (middle), and fluid ingested in the form of "drinks" (bottom) in the bouts (5 hatched bars), in the total intakes over the day (solid bar) and over the week (cross-hatched bar). The first bar of each set of five hatched bars representing the intake in bouts represents the bout definition of minimum 15 minute IMI and 50 ml size; the second, 45 rain/50 ml; the third, 45 rain/100 ml; the fourth, 45 min/200 ml; and the fifth, 90 min/50 ml. *Indicates that the mean is significantly (p<0.05) different from zero as assessed with a t-test.
The correlations calculated between the nutrient and the fluid intakes for individual bouts are represented by the hatched bars in Fig. 1. The pattern is very similar to that found for intakes over the entire day. Carbohydrate again has the strongest relationship with fluid intake compared to fat [t(218)=5.80, 5.78, 8.51, p
516
~÷ CASTRO
Total ,5
CorroJatllon$
Gluid
T * *
$
,
,
.3
.1
o
Z
gz -.2
Fluid • u
* * ,
,
o
.3
Carbohydrato
Fat
Protllm
;odi um
o"
u
.i
-.i
Multiple
Regreosiono
÷
g
-.3
1 Orink .5 .4
Pluid onl X [] MeQl Int~H~ Pr~dictlon I Doily Intake Prediction [] Weekly I n t o ~ Prediction
.3 * *
* , ,
.!
-.| -.2
Carbohydrate
Carbohydrat~
Fat
Protein
Fat
Pro~eim
Sodzum
5od~ um
FIG. 2. Mean_ SEM beta coefficients from the multiple linear regressions with coingested carbohydrate (left 7 bars), fat (middle left 7 bars), protein (middle right 7 bars), and sodium (right 7 bars) as independent variables predicting the dependent variables of the amounts of total fluid (top), fluid in excess of digestive requirements (middle), and fluid ingested in the form of "drinks" (bottom) in the bouts (5 hatched bars), in the total intakes over the day (solid bar) and over the week (cross-hatched bar). The first bar of each set of five hatched bars representing the intake in bouts represents the bout definition of minimum 15 minute 1MI and 50 ml size; the second, 45 min/50 mi; the third, 45 min/100 mi; the fourth, 45 rain/200 ml; and the fifth, 90 min/50 ml. *Indicates that the mean is significantly (p<0.05) different from zero as assessed with a t-test. The multiple regressions calculated between the nutrient and the fluid intakes over the entire day are represented by the solid (filled) bars in Fig. 2. Although the magnitude of the beta coefficients varies according to whether total fluid (top), excess fluid (middle) or " d r i n k " fluid (bottom) is used for the dependent variable, once again, as with the correlations, the same pattern emerges. Carbohydrate has the strongest relationship w i t h fluid intake compared to fat [t(218) = 3.14, 2.88, 2.84, p<0.05], compared to protein [t(218)=2.16, 1.28 (n.s.), 3.16, p<0.05], and compared to sodium [t(218)=2.51, 1.26 (n.s.), 2.76, p<0.05] for total, excess, and "&'ink" fluid, respectively. On the other hand fat has the weakest relationship with fluid intake compared to protein [t(218)=2.68, 2.75, 1.36 (n.s.), p<0.05] and compared to sodium [t(218)=2.54, 2.96, 1.70 (n.s.), p<0.05] for total, excess, and " d r i n k " fluid, respectively. In no case do protein and sodium significantly differ, The beta coefficients calculated over the entire weeks intakes are represented by crosshatched bars in Fig. 2. The pattern for these coefficients is very similar to that for the daily intakes except that the carbohydrate beta coefficients are considerably smaller while for protein they are larger.
FIG. 3. Mean± SEM correlations (top) and beta coefficients from the multiple linear regressions (bottom) between coingested carbohydrate (left 5 bars), fat (middle left 5 bars), protein (middle right 5 bars), and sodium (right 5 bars) and the amount of time until the next bout containing fluid. The first bar of each set of five represents the bout definition of minimum 15 minute IMI and 50 ml size; the second, 45 min/50 ml; the thirdi 45 min/100 mi; the fourth, 45 min/200 ml; and the fifth, 90 min/50 ml. *Indicates that the mean is significantly (p<0.05) different from zero as assessed with a t-test.
The beta coefficients from the multiple regressions calculated between the nutrient and the fluid intakes for individual bouts are represented by the hatched bars in Fig. 2. The pattern is very similar to that found for intakes over the entire day and for the univariate correlational analyses. Carbohydrate again has the strongest relationship with fluid intake compared to fat [t(218)= 12.71, 12.92, 11.94, p<0.05], compared to protein [t(218)= 1.32 (n.s.), 1.52 (n.s.), 7.12, p<0.05], and compared to sodium [t(218) = 4.52, 3.97, 11.53, p<0.05] for total, excess, and "drink" fluid, respectively. Again fat has the weakest relationship with fluid intake compared to protein [t(218)= 11.96, 12.23. 4.21, p<0.05] and compared to sodium [t(218)=7.54, 8.27. 7.12. p<0.05] for total, excess, and " d r i n k " fluid, respectively. Protein has significantly stronger correlations than sodium [t(218)= 5.47, 5.14, 3.79, p<0.05] for total, excess, and " d r i n k " fluid. respectively. The macronutrients and sodium ingested in the bouts were also used to predict the amount of time that would elapse before the next occasion of fluid ingestion (after bout interval). The univariate correlations (top) and the beta coefficients from the multiple linear regression analyses (bottom) are presented in Fig. 3. The macronutrients and sodium all have small but significant positive correlations with the after bout interval. There were no
NUTRIENTS AND FLUID INTAKE
Poeebout
$elf-Roted
517
T~lr~t
Pre
Correlotto~a 0
-.1
to
Poe¢bout
Chnnge i n $mlF-Ru~md
Thiret
Cerrelatlor~
g
.3
1 ii
7, z E
o~ 0
z z z 7, z
.2
0
=[
°
.
g
I
-.3
Mul t l p l e RegreaeJone
Multlpl@ RRgr@m@Ion@
tJ
o
*
,
o~
I -.2
Ccn-t~ol~ya~ate
Foe
ProgeJn
$odltffe
TOrn]
Wnt~
FIG. 4. Mean +-SEM correlations (top) and beta coefficients from the multiple linear regressions (bottom) between coingested carbohydrate (left 5 bars), fat (middle left 5 bars), protein (middle 5 bars), sodium (middle right 5 bars), and total water (fight 5 bars) and the self-ratings of subjective thirst at the end of the bout. The first bar of each set of five represents the bout definition of minimum 15 minute IMI and 50 ml size; the second, 45 rain/50 ml; the third, 45 rain/100 ml; the fourth, 45 rain/200 ml; and the fifth, 90 rain/50 mi. *Indicates that the mean is significantly (,o<0.05) different from zero as assessed with a t-test. significant differences between the magnitudes of these correlations. Similarly, the beta coefficients from the multiple regressions predicting the after bout interval were all small and positive but only rarely significant. There were also no significant differences between the magnitudes of these beta coefficients. The macronutrients, sodium and total water ingested in the bouts were also used to predict the level of self-rated thirst at the end of the bout. The univariate correlations (top) and the beta coefficients from the multiple linear regression analyses (bottom) are presented in Fig. 4. The nutrients, sodium, and total water all have significant negative correlations with postbout self-rated thirst, the more of each the lower the thirst. Protein had significantly larger correlations than any of the other constituents of the bouts [t(129) = 3.07, 2.10, 1,99, 2.53, p<0.05] for carbohydrate, fat, sodium and total water, respectively. There were no significant differences between the correlations obtained for any of the other constituents. The beta coefficients from the multiple regressions also indicated negative relationships between the constituents and thirst but only in the cases of carbohydrate, protein and total water were the beta coefficients significantly different from zero. These results indicate the more carbohydrate, protein, or total water ingested in a bout the lower the thirst self-rating at the end of the bout. These same factors were used as predictors of the change in
-. 1
FIG. 5. Mean__.SEM correlations (top) and beta coefficients from the multiple linear regressions (bottom) between coingested carbohydrate (left 5 bars), fat (middle left 5 bars), protein (middle 5 bars), sodium (middle fight 5 bars), and total water (fight 5 bars) and the change in the self-ratings of subjective thirst from prior to the bout to the end of the bout. The f'wst bar of each set of five represents the bout definition of minimum 15 minute IMI and 50 ml size; the second, 45 rain/50 ml; the third, 45 rain/ 100 ml; the fourth, 45 min/200 ml; and the fifth, 90 rain/50 mi. *Indicates that the mean is significantly (p
self-rated thirst from the start of the bout to the end. The unlvariate correlations (top) and the beta coefficients from the multiple linear regression analyses (bottom) are presented in Fig. 5. The nutrients, sodium, and total water all have significant positive correlations with the change in thirst, the more of each the greater the reduction in thirst occurring during the bout. The only significant difference between these correlations was that the correlation with fat intake was significantly lower than that with either protein, sodium, or total water [t(129)=2.61, 2.21, 2.14, p<0.05], respectively. The multiple regression analysis produced beta coefficients which were positive and significantly different from zero for carbohydrate, protein, and total water, indicating that the more of these constituents present the greater the reduction in thirst occuring over the course of the bout. The beta coefficients for fat significantly differed from carbohydrate, protein, and total water [t(129)= 2.79, 2.26, 3.81, p<0.05], respectively. DISCUSSION The present results indicate that the fluid ingested during bouts or over the course of a day or a week is not primarily related to the depleting/repleting characteristics of the coingested nutrients. The amount of time elapsing until the next bout, also, does not
518
appear to be influenced by the nutrients' impact on fluid balance. Additionally, the subjective state of thirst, although greatly affected by the ingestion of nutrients, does not appear to be related, either, to the nutrients' impact on body fluids. These results, then, suggest that fluid intake is not anticipatory of the fluid depleting/ repleting impact of the digestion and metabolism of ingested food. They further suggest that under ad lib conditions in the natural human environment the intake of fluids and the subjective state of thirst would not appear to be primarily determined by fluid homeostasis. The expected large impact of sodium ingestion was not present. Although sodium intake correlates with fluid ingestion, this correlation would appear to result from the covariation of sodium intake with the overall meal size. When sodium intake was used in a multiple regression along with the three macronutrients its relationship with either total fluid or excess fluid ingested became small and with "drink" fluid vanished entirely. This was true regardless of whether the intakes occurring in bouts, over the entire day, or even over the entire week were used in the calculations. Additionally, it was expected that if the influence of ingested sodium on the effective plasma osmolality was an influence on the regulation of fluid intake then a negative relationship would be expected between sodium intake and the after bout interval. It was expected that the more sodium ingested in the bout the shorter the interval until fluid was again ingested. In fact, the opposite was found with the amount of sodium in the bout positively associated with the duration of the after bout interval. Sodium ingestion also did not have the expected effect on the subjective state of thirst. It would be expected that sodium would be positively correlated with thirst and negatively related to the satisfaction of thirst, i.e., the change scores. In fact, the opposite was found with the amount of sodium ingested negatively correlated with postbout thirst and positively correlated with the change in thirst over the bout. Indeed, even when the influence of the amount of water ingested in the bout was removed in the multiple regression, sodium intake failed to show the expected influence on the subjective state. Thus neither the amount of fluid ingested, its timing, nor the subjective state of thirst or its relief would appear to be strongly influenced by ingested sodium. Thus it does not appear that thirst and drinking are anticipatory to the postingestive effects of sodium on the effective plasma osmolality (35) whether looked at over meal, daily, or weekly time spans. These results, then, call into question the importance of plasma osmolality as a major determinant of either ad lib fluid intake or the subjective state of thirst. Protein intake, since it produces a relatively large turnover of water, was expected to be strongly related to fluid intake as has been observed in manipulative experimental studies (25). To a moderate extent, the expected relationships with fluid ingestion were found but not with the timing of intake or the subjective state of thirst. Protein intake was found to be positively related to total and excess fluid intake ingested in bouts, over the day, and over the week with both univariate and multivariate analyses, but only weakly related with fluids ingested in the form of "drinks." However, in many ways protein ingestion was not found to have the expected relationships. It was expected that since protein would have a fluid-depleting effect, then the more protein ingested in the bout the shorter the interval until fluid was again ingested. This was not the case. Indeed, protein ingestion was positively related to the duration of the postbout interval. Additionally, protein ingestion would be expected to be associated with higher postbout thirst and a smaller change in subjective thirst over the bout. In fact the opposite was found. Protein intake not only had a negative association with postbout thirst but, in fact, had the strongest negative association of all of the constituents looked at. Thus, once again, the present results do not
~k CASTRO
support the notion that anticipation of changes in body fluid homeostasis is the determining factor in the regulation of fluid mtake and subjective thirst. Carbohydrate intake, since it has only a small osmotic impact (34), was not expected to be strongly related to fluid intake or to the subjective state of thirst. Contrary to this prediction, carbohydrate, of all the measured constituents, was probably the most strongly related to fluid intake and was among the strongest in association with the postbout interval and the subjective state. The only constituent that did have the expected relationships with fluid intake and thirst was fat. Based upon the fact that fat ingestion does not have osmotic effects, it would not be expected to b¢ related to either fluid intake or thirst. Indeed, fat intake was found to have little discernable association with either the amount of fluid ingested, the timing of fluid intake, or the subjective state of thirst at the end of the bout or its change over the bout. The present results, then, clearly do not support an explanation of fluid intake based upon the anticipation of changes in body fluid balance resulting from ingested nutrients. Intake and subjective state were not related to the osmotic or volumetric properties of the coingesta. What, then, determines fluid intake and thirst? Clearly, the foods normally ingested by humans in their normal everyday existences are highly hydrated and in combination with "drinks" taken along with the foods produce a marked surfeit of fluids over that needed for digestion. This is evidenced by the fact that over 1.5 liters of water per day, over 80% of the total daily amount of fluid, was ingested in excess of digestive requirements (Table 1). Since, the nutrients are so highly hydrated, it would seem reasonable that their ingestion would be positively related to the satisfaction of subjective thirst. Additionally, this high degree of hydration would mean that any depleting effects of the nutrients would be more than compensated for. The results make more sense when conceptualized in reference to the fluid repletion effects of the bout rather than depletion effects. Since fat and sodium are normally not hydrated, the fact that they were not related to thirst and fluid intake makes sense, given they are not contributing any fluid to the system. Carbohydrate and protein in the human diet, on the other hand, are heavily hydrated and thus contribute to repletion rather than depletion. Thus the positive relationships of these two nutrients with fluid intake and thirst reduction are readily understandable. In prior papers, it was suggested that fluid intake was mainly governed by feeding in the ad lib pattern of intake of both rats and humans (7, 9, 10). Fluid was seen as ingested well above requirements and the job of maintaining body fluid balance was left to the kidney. The present results support this interpretation. In addition, the present results suggest that fluid intake may be primarily governed by palatability factors, with both the palatability of the "drinks" and the solid foods involved. The clear and strong relationship of carbohydrate with the amount of fluid ingested suggests that sweetness may be an important determinant. Since ingested "drinks" frequently contain carbohydrate and since ingested fluids increase the palatability of ingested foods, palatability effects alone could account for the covariation of food and fluid intake in the natural ad lib pattern of ingestive behaviors. Self-reports of food intake have traditionally been thought to be inaccurate. In the case of diet recall procedures there probably is considerable inaccuracy (24). However, the diet diary technique where subjects record at the same time that they eat has been demonstrated to be both reliable and valid (1, 21, 24, 30). Additionally, in the present study the subjects' reward for accurate record keeping was a detailed analysis of their reported di, ets. The subjects expressed great interest in receiving such an analysis and were aware that it would only be as accurate as their records. Also, the subject's diary entries were verified by two people who ate with the subject. Thus there is every, reason to
NUTRIENTS AND FLUID INTAKE
519
believe that the self-reports acquired in the present study are accurate. Even if the technique produces many errors, there is no reason to suspect any systematic relationship between recording errors and the primary variables of interest in the study. Unsystematic error should cover up significant relationships not produce them. The fact that clear, replicable, and highly significant results, as are reported in the present study, can be discerned with a somewhat insensitive and error prone technique speaks to the power and importance of the effects reported. The present results clearly indicate that under ad lib conditions in humans fluid intake and subjective thirst are not related to the depleting characteristics of the ingested nutrients. Rather, fluid intake and thirst are related to their repleting characteristics. This again supports a model that ad lib fluid intake occurs in response to and as an adjunct of food intake, that as a result fluid is ingested in excess of requirements, and that the kidney is primarily responsible for body fluid homeostasis. This interpretation does not necessarily apply to other ecological conditions. The data for
the present study were obtained from primarily middle class Americans living in a m o d e m urban environment where food and fluids are readily and abundantly available. In situations where fluids are not available ad lib or where obtaining water might be difficult or dangerous, tighter regulation of fluid intake independent of food intake may well occur (3,29).
ACKNOWLEDGEMENTS The author would like to express his appreciation and acknowledge the substantial contributions of Ms. Dixie K. Elmore, Marie Brewer, and Ms. Margaret Pedersen without whose assistance the work could not have been performed, and to Dr. Terry Thrasher and Dr. Gary Robertson for suggesting the analysis. This research was supported in part by Grant R01-DK39881-01A2 from the National Institute of Diabetes and Digestive and Kidney Diseases, from a grant from the Georgia State University Research Grant Program and from a Biological Research Support Grant whose support is gratefully acknowledged.
REFERENCES 1. Adleson, S. F. Some problems in collecting dietary data from individuals. J. Am. Diet. Assoc. 36:453-461; 1960. 2. Bigelow, J. E.; Houpt, T. R. Feeding and drinking patterns in young pigs. Physiol. Behav. 43:99-109; 1988. 3. Collier, G. The dialog between the house economist and the resident physiologist. Nutr. Behav. 3:9-26; 1986. 4. de Castro, J. M. Meal pattern correlations: Facts and artifacts. Physiol. Behav. 15:13-15; 1975. 5. de Castro, J. M. Macronutrient relationships with meal patterns and mood in the spontaneous feeding behavior of humans. Physiol. Behav. 39:561-569; 1987. 6. de Castro, J. M. Circadian rhythms of the spontaneous meal patterns, macronutrient intake, and mood of humans. Physiol. Behav. 40:437466; 1987. 7. de Castro, J. M. A microregulatory analysis of spontaneous fluid intake by humans: Evidence that the amount of liquid ingested and its timing is mainly governed by feeding. Physiol. Behav. 43:705-714; 1988. 8. de Castro, J. M. Physiological, environmental, and subjective determinants of food intake in humans: A meal pattern analysis. Physiol. Behav. 44:651-659; 1988. 9. de Castro, J. M. The interaction of fluid and food intake in the spontaneous feeding and drinking patterns of rats. Physiol. Behav. 45: 861-870; 1989. 10. de Castro, J. M. Social, circadian, nutritional, and subjective correlates of the spontaneous pattern of moderate alcohol intake of normal humans. Pharmacol. Biochem. Behav. 35:923-931; 1990. 11. de Castro, J, M. Bout pattern analysis of ad lib fluid intake. In: Ramsay, D. J.; Booth, D. A., eds. Thirst: Physiological and psychological aspects. Bedim Springer-Verlag; 1991:in press. 12. de Castro, J. M.; Brewer, M.; Elmore, D. K.; Orozco, S. Social facilitation of the spontaneous meal size of humans is independent of time, place, alcohol, or snacks. Appetite 15:89-101; 1990. 13. de Castro, J. M.; de Castro, E. S. Spontaneous meal patterns of humans: influence of the presence of other people. Am. J. Clin. Nutr. 50:237-247; 1989. 14. de Castro, J. M.; Elmore, D. K. Subjective hunger relationships with meal patterns in the spontaneous feeding behavior of humans: Evidence for a causal connection. Physiol. Behav. 43:159-165; 1988. 15. de Castro, J. M.; McCormick, J.; Pedersen, M.; Kreitzman, S. N. Spontaneous human meal patterns are related to preprandial factors regardless of natural environmental constraints. Physiol. Behav. 38: 25-29; 1986. 16. Engell, D. Interdependency of food and water intake in humans. Appetite 10:133-141; 1988. 17. Falk, J. L. Production of polydipsia in normal rats by an intermittent
food schedule. Science 133:195-196; 1966. 18. Fitzsimons, J. T. Drinking by nephrectomized rats injected with various substances. J. Physiol. (Lond.) 155:563-579; 1961. 19. Fitzsimons, J. T. Drinking of rats depleted of body fluid without increase in osmotic pressure, J. Physiol. (Lond.) 159:297-309; 1961. 20. Fitzsimons, J. T.; LeMagnen, J. Eating as a regulatory control of drinking. J. Comp. Physiol. Psychol. 67:273-283; 1969. 21. Gersovitz, M.; Madden, J. P.; Smicikalas-Wright, H. Validity of the 24-hour dietary recall and seven-day record for group comparisons. J. Am. Diet. Assoc. 73:48-55; 1978. 22. Gilman, A. The relation between blood osmotic pressure, fluid distribution and voluntary water intake. Am. J. Physiol. 120:323-328; 1937. 23. Kissileff, H. R. Food associated drinking in the rat. J. Comp. Physiol. Psychol. 67:284-300; 1969. 24. Krantzler, N. J.; Mullen, B. J.; Schultz, H. G.; Grivetti, L. E.; Holden, C. A.; Meiselman, H. L. The validity of telephoned diet recalls and records for assessment of individual food intake. Am. J. Clin. Nutr. 36:1234-1242; 1982. 25. Le Magnen, J.; Tallon, S. Les EMterminants quantitatifs de la prise hydrique clans ses relations avec la prise d'aliments chez le rat. C R Soc. Biol. (Paris) 161:1243-1246; 1967. 26. Lucas, G. A.; Timberlake, W.; Gawley, D. J. Learning and meaiassociated drinking: meal-related deficits produce adjustments in postprandial drinking. Physiol. Behav. 46:361-367; 1989. 27. Phillips, P. A.; Rolls, B. J.; Ledingham, J. G. G.; Morton, J. J. Body fluid changes, thirst and drinking in man during free access to water. Physiol. Behav. 33:357-363; 1984. 28. Rolls, B. J.; Wood, R. J.; Stevens, R. M. Effects of palatability on body fluid homeostasis. Physiol. Behav. 20:15-19; 1978. 29. Rowland, N. Regulatory drinking: do physiological substrates have an ecological niche? Biobehav. Rev. 1:261-272; 1977. 30. St. Jeor, S. T.; Guthrie, H. A.; Jones, M. B. Variability of nutrient intake in a 28 day period. J. Am. Diet. Assoc. 83:155-162; 1983. 31. Striker, E. M. Some physiological and motivational properties of the hypovolemic stimulus for thirst. Physiol. Behav. 3:379-385; 1968. 32. Toates, F. M. A physiological control theory of the hunger-thirst interaction. In: Booth, D. A., ed. Hunger models. New York: Academic Press; 1978:347-374. 33. Toates, F. M. Computer simulation and the homeostatic control of behaviour. In: McFarland, D. J., ed. Motivational control systems analysis. New York: Academic Press; 1974:407--426. 34. Voices, T. Water homeostasis. Annu. Rev. Nutr. 7:383---406; 1987. 35. Zerbe, R. L.; Robertson, G. L. Osmoregulation of thirst and vasopressin secretion in human subjects: effects of various solutes. Am. J. Physiol. 244:E607-E614; 1983.