Meal pattern analysis: artifacts, assumptions and implications

0361~923wM $3.00+ ml

Brain Research RuMin, Vol. 17, pp. 439-443, 1986.@Ankho InternationalInc. Printedin the U.S.A.

Meal Pattern Analysis: Artifacts, Assumptions and Implications’” THOMAS Food Intake

Laboratory

W. CASTONGUAY, and Department

LUCIA of Nutrition,

L. KAISER University

AND JUDITH S. STERN of California-Davis,

Davis, CA 95616

CASTONGUAY, T. W., L. L. MISER AND J. S. STERN. Meal pattern analysis: Artifacts, assumptions and impiications. BRAIN RBS BULL 17(3) 439-443, 1986.-The meal patterns of 12 male Sprague-Dpwky rats were monitored continuously for eight consecutive days. During that time, food intake was measured every second, accurate to 0.01 gram. Results from this procedure demonstrated that the comlation between meal size and meal duration was, at best, weak. Further, the correlation between meal size and either the pn or post meal interval was also weak (approximately 0.20). Subsiquent re-evaluation df the patterns using different end-of-the-meal. criteria resulted in a @ifkant interaction betweentheend~f-the-meal&~~andthesbnagthoftbecomhrtioabetwanmeal~aadpwtmealiaterval,~~ robust correlations beii observed with the use of longer end-of-the-meal definitions. In an attempt to resolve the question of which definition to use, log survivorship analysis was applied to the interval data. Results from that procedure suggest that a 10 minute end-of-the-meal definition is appropriate in most cases in the analysis of daytime patterns, and a 5 minute ddition should be used when quantifying nighttime patterns. Under these “data determined” conditions, the correlation between meal size and post meal interval was not statistically significant. The implicati
End-of-the-meal criteria

given access to the same composite diet, the laboratory rat eats consistent amounts of food from day to day [ 11. Further, if the diet is then diluted with non-caloric materials, the rat will often adjust its intake so as to consume the same daily caloric load as it had prior to dietary dilution. Over the past 50 years, researchers have sought to uncover the physiological and behavioml mechansims that permit this consistency, often referred to as “caloric regulation” [ 1, 6, 111. One of the most attractive regulatory hypotheses that has been examined is that the brain is actively monitoring the presence of some factor derived from each meal [7]. AU subsequent feeding is then adjusted so as to maintain some privileged level of the factor. Experimental evidence supporting this hypothesis has been derived from the study of ad lib feeding patterns. It follows from the hypothesis that if food intake is regulated from meal to meal, then the intervals after meals shoukl be correlated with the size of any given meal. Several investigators have observed significant correlations between meal size and post meal intervals [4,7, 153. Others have failed to observe these relations [3, 10, 121. Because of the increased precision of modem electronics as well as the relative ease of computerized meal pattern data analysis, we sought to reexamine the relations between not only meal size and intermeal intervals, but also the effect of’variations in a parameter that is used to defme the end of a meal [2]. WHEN

METHOD

Animals and Apparatus Twelve male Sprague-Dawley rats (Simonsen Labs, Gilroy, CA), weighing 150-160 g were individually housed in cages that were equipped to monitor ad lib meal patterns. Each cage (or “meal pattern module”) was equipped with a Mettler P300 balance, on which was placed a food cup. An LSI 11102 computer, located in an adjacent room, monitored the output from the balances, updating its reading every second. In that way, second by second recordings were made of changes in the output of each balance. A custom made food cup was placed beneath an opening in the side of each cage. The cup was designed so that a spillage lip surrounded the opening in the cup. In that way any spillage that dropped from the paws or face of the feeding rat was likely to fall either back into the cup or onto the lip of the same cup. All animals were housed in one laboratory room that was maintained at 2222°C. Procedure Each rat was placed in a meal pattern module and allowed to feed ad lib on Purina Rat Chow and water. The. rats were

allowed to adapt to the experimental

cages for 7 days, after

‘This research was supported in part by Postdoctoral Training Grant No. T32 AM 07355 and Grant AM 18899 from the National Institutes of Health. Portions of this work were presented at the Mechanisms of Appetite and Obesity symposium held in San Antonio, TX and the 15th Annual Meetings of the Society for Neuroscience in Dallas, TX. “Requests for reprints should be addressed to Thomas W. Castonguay, Nutrition Department, University of California-Davis, Davis, CA 95616.

439

440

CASTONGUAY, TABLE

1

2

AVERAGE MEAL SIZE OF SPRAGUE-DAWLEY RATS lnr- II AS A FUNCTION OF END-OF-THE-MEAL DEFINITIONS

definition (minutes) 10

5

~

20

Meal size (g)

~~~~

End-of-the-meal

40

Daytime

3.51 to.34

3.42 kO.32

3.37 to.39

3.40 k-o.31

3.25 +0.22

Nighttime

10.62 20.16

9.10 r0.52

7.99 kO.42

6.50 20.27

3.77 fO.22

which meal patterns were collected and analyzed for another 8 consecutive days. Each day at 0900 hr throughout the adaptation and data collection phases, each rat was weighed, and its water bottle and food cup refilled. Entrance into the animal room was restricted to these maintenance sessions. One 15 watt incandescent lamp was used as a house light for each of the 12 modules. These house lights were maintained on a 12 hr/l2 hr day-night cycle, with lights coming on at 0900 hr. In that way, maintenance was scheduled to occur at the beginning of the rats quiescent period.

As with any meal pattern analysis, two parameters had to be determined before data reduction could occur. First, the minimum interval that was required to have elapsed after a feeding session was experimentally investigated. Five such intervals, or end-of-the-meal definitions, were investigated: 2 min. 5 min. IO min, 20 min, or 40 min. The second parameter was the minimum intake required to define the start of the meal. Previous work [lo] had suggested a minimum meal size definition of 0.05 g was sufficiently sensitive to include all feeding episodes without introducing experimental error, such as balance drift. Daytime and nighttime meal frequency and average meal size from the meal pattern analyses were used as dependent measures and were analyzed with the use of a repeated measures analysis of variance. Pearson Product-Moment correlations coefficients were computed individually for each rat’s intermeal interval, meal size, and meal duration data [7]. RESULTS

Frrquency

und Size

The average meal frequency and average meal size data for both the daytime and nighttime are presented in Tables I and 2. Note that as the end of the meal definition was increased, meal frequency during the nighttime. but not the daytime declined (daytime meal frequency ANOVA Detinition effect F(4,44)= I .54, p>O. IO; nighttime meal frequency ANOVA Definition effect F(4,44)=54.61: p
results from the present set of

definition (minutes)

2

5

IO

20

40

Daytime

1.14 20.08

1.16 ?0.08

1.23 to.09

1.3X kO.14

1.79 kO.19

Nighttime

1.42 kO.09

1.63 to.09

1.84 co. 11

2.16

_to.I I

3.31 -co.17

Note the dramatic decline in nighttime meal frequency with increasing end-of-the-meal definitions. All values are means t SEM.

Meal

AND STERN

TABLE 2

MEAL FREQUENCY OF SPRAGUE-DAWLEY RATS (II= 12) AS A FUNCTION OF END-OF-THE-MEAL DEFINITIONS

End-of-the-meal

KAISEK

Note the dramatic increase in average meal size found by increasing the end-of-the-meal definition. especially during the night. All values are means ‘_’SEM.

TABLE 3 MEAL DURATION VS. MEAL SIZE CORRELATIONS IN SPRAGUEDAWLEY RATS AS A FUNCTION OF THE END-OF-THE-MEAL DEFINITION

End-of-the-meal

definition (minutes)

2

s

IO

20

40

Daytime

0.825 -to.044

0.742 kO.09

0.682 +o.os

0.582 tO.OX

0.624 to.06

Nighttime

0.588 kO.078

0.807 to.036

0.771 50.032

0.7 t I -c0.051

0.814 kO.021

All values are means t SEM. Note that shorter end-of-the-meal definitions promote higher meal duration/size correlations in the daytime. That was not found to be the case within the nighttime data. Each mean in the above table was calculated by averaging the individual r value from each rat’s meal pattern data. The data used for these analyses was recorded from 12 rats monitored over 8 consecutive days. All of the above correlation coeffkients are statistically significant at or beyond the 0.01 confidence limit.

analyses grew out of the opportunity to correlate meal size with meal duration. Many investigators have relied upon meal duration as a reliable correlate of meal size. Presented in Table 3 are the average daytime and nighttime Pearson Produce-Moment correlation coefficients for the I? rats over the 7 day monitoring period for each of the 5 end of the meal definitions. The range of values during the daytime was from -0.126 (Rat No. 8, 5 minute end-of-the-meal definition) to 0.971 (Rat No. 2, 5 minute end-of-the-meal definition). The range of the values during the nighttime was from 0.223 (Rat No. 11, 2 minute end-of-the-meal definition) to 0.979 (Rat No. 4, 2 minute end-of-the-meal definition). In general. the relationship between meal size and duration is not as strong as would be needed to accurately estimate meal size from duration data.

The relationships between the intervals before and after each meal with meal size were very weak. Listed in Table 4 are the Pearson correlation coefficients for daytime nighttime meals. Of particular note is the relationship

and be-

MEAL

PATTERN

ANALYSIS

441

TABLE INTERMEAL

4

INTERVAI.MFiAL.

SIZE CORRELATIONS FOR SPRAWJE-DAWLEY AS A FUNCI’IONOF END-OF-THE-MEAL DEFINITIONS

End of the meal deftition (mW

RATS

(n= 12)

Daytime

Nighttime

Mean rt SEM

Mean f SEM

Premeal interval vs. meal size

2 5 10 20 40

-0.098 -0.122 -0.097 -0.246 -0.248

+ -’ 2 f f

0.050 0.048 0.067 0.050 0.050

Postmeal interval vs. meal size

2 5 10 20 40

0.094 0.063 -0.037 -0.134 -0.143

2 + + + f

0.097 0.101 0.106 0.099 0.0699

0.212* 0.184* 0.154 0.134 0.029

+ f 2 * +

0.029 0.036 0.046 0.045 0.072

-0.047 f 0.050 -0.028 + 0.041 0.024 f 0.044 0.144 2 0.048 0.288 f 0.071

All values are means f SEM. Each mean was calculated by taking the average value of individual r values from each rat’s data. Data were collected over a period of 8 consecutive days. Note that increases in the end-of-the-meal definition promote stronger correlations between post meal interval and meal size during the nighttime. Coefficients denoted with an asterisk (*) are statistically significant at or beyond the 0.05 confidence limit.

soo- .

L-OF

aMn.AnvE-SURVlWWj 1.00 .80

TOO-

5

60 A0 20

%wJwoti

.I0 1-

:

600-

j

500:

g u.

400-

2 z it

300-

z

200-

l

.

.. .

.

. .

. .

.

ioo-

200

FIG. I. An illustrative example of a Log Survivorship function. Plotted on the abscissa is the cumulative proportion of all of the intervals between feeding episodes that were longer than one second in duration that remain (that survive) after eliminating the values less than or equal to the value of the ordinate. Each function’s breakpoint was defined as the point on the curve at which the slope of the log survivor curve initially changed.

st meal interval/meal size correlation when variations in tR”e end of the meal definition are performed. The tween the

statistical relationship between the two measures is greatly strengthened with longer end-of-the-meal definitions. Log Survivorship

Analysis

Because of the effects of variations in the end-of-the-meal definitions, we set out to ask the question “Is there a biologically relevant end of the meal definition?” To address that question, a second set of analyses were performed. Log Survivorship Analysis [14] was applied to each rat’s set of

600

1000 I400 MOO 2200 2600 3000 3400 3600 4200 BREAK POINT (seconds)

FIG. 2. Breakpoints were calculated for each rat using all of the interbout interval data. Note that as the number of intervals between feeding bouts increased, the breakpoint decreased.

intermeal intervals [ 12,131. To perform that analysis, the end: of-the-meal definition was first set to 1 second. In that way, all of the intervals between changes in balance output (both within meal pauses and between meal pauses) were used. A “breakpoint” in the function was then determined as the first incidence of a significant change in the slope of the exponential function. An example of the log survivorship analysis is presented in Fig. 1. Presented in Fig. 2 is a plot of each rat’s breakpoint presented as a function of the number of intervals used to determine breakpoint. Each breakpoint was calculated by using all of the intervals between feeding episodes for the entire 7 day monitoring period. Note that the relationship between breakpoint and interval frequency is highly corre-

CASTONGUAY.

442

FIG. 3. Separate daytime and nighttime breakpoints were calculated for each rat. Nighttime breakpoints were much shorter than were those calculated from daytime data.

lated (r= -0.753, ~~0.01). Although the average breakpoint was 23.9 minutes (1434 seconds), if only those animals having more than 400 intervals were used, the average breakpoint was 10 minutes, 25 second (625 seconds). The correlation between the number of data points used in the log survivorship analysis and the resultant “breakpoint” was nevertheless significantly correlated (-0.751). Because of the large difference in meal frequency between daytime feeding and nighttime feeding, separate daytime and nighttime log survivorship functions for each rat were calculated. The results from these analyses are presented in Fig. 3. Two very different distributions were found, with breakpoints of nighttime meals found uniformly between 250 and 450 seconds, and daytime breakpoints found between 400 and 2500 seconds. From this second set of analyses, it seems clear that the variability found observed in the combined data was principally attributable to the daytime observations. DISCUSSION

We have already noted that variations in the definitions of both the end-of-the-meal as well as minimum meal size can sometimes dramatically alter the conclusions that are drawn about meal frequency and meal size results 121. The present results not only replicate that original set of observations. but extend their applicability to include any conclusions that are drawn about the relationship between meal size and the intermeal interval. Most investigators have reported correlations between meal size and intermeal interval using an “arbitrarily” chosen end-of-the-meal defmition ([3.7. 10. 12. 151; but see [2,4]). From the present results, it seems clear

KAlSER

AND STERN

that the strength of that correlation is in part determined by the choice of that definition. The importance of the correlation is rooted in the argument that was first advanced by Le Magnen and colleagues [7-91. Namely, under ad lib conditions the energy derived from a meal is actively monitored by the brain. During the post ingestive period, this energy is used as a function oi current metabolic rate. When these meal-derived stores drop below some threshold, the animal will initiate another bout of feeding. If this theory is correct, then one would predict that a strong relationship exists between the energy derived from any one meal and the interval subsequent to that meal. In support of the theory, several investigators have reported statistically significant correlations between meal size and post meal interval. Other investigators have failed to obtain significant correlations [ 10,121. Further, they have noted that the reported significant correlations often fell far from accounting for much of the variability between two factors. The present results not only fail to support the homeostatic theory, but also suggest a plausible explanation IO account for those reports of a significant correlation between postmeal interval and meal size. The use of long end-ofthe-meal definitions significantly contribures to the strength of the correlation between meal size and post meal interval. Since Le Magnen and colleagues used long end-of-the-meal definitions, it is not surprising that robust correlations were found. We have gone on to suggest. however. that when log survivorship analyses are applied to the intcrbout interval data the definitions for the end of a meal are usually short. between 5 and 10 minutes. If one adopts these criteria for the end-of-the-meal definition. then the correlation between meal size and post meal interval is not different from zero: I.e., there is only a random relationship bctwecn the IWO factors. The implications of these findings is that a depletionrepletion model of feeding is probably overly simplified to account for meal taking behavior. The most serious limitation of the present set of analyses is that they fail to provide any further resolution to the question of how the rat adjust\ its intake of food so as to meet metabolic demands. One final issue that deserves comment from these analyses concerns the relationship between the duration of feeding bouts and meal size. The suggestion from the present experiment is that meal duration. while a correlate of meal size, is not a good substitute for measuring meal Gre.

ACKNOWLIiDGtiMIiNIS

The authors would like to thank Jamr\ Gihh\. M.D. of the Bourne Research Laboratory for hi\ insightful comments on an earlier presentation of this work. The authors would also like to express their appreciation to Mr. Jean Bernier. Ms. Christina Wu Nasrawi and Mr. Glen Bergman for their technical assistance.

REFERENCES Adolph, E. F. Urges to eat and drink in rats. Am J Physiol 151: 110-125, 1947. Castonguay, T. W., D. E. Upton, P. M. B. Leung and J. S. Stem. Meal patterns in the genetically obese Zucker rat: A rcexamination. Physiol Behav 28: 911-916, 1982. Collier, G. H., E. Hirsch and P. Hamhn. The ecological determinants of reinforcement in the rat. Physiol Behtrl. 9: 399-410. 1972.

4. Danguir, J., S. Nicolaidis and H. Gerard. Relations between feeding and sleep patterns in the rat. .I C,vr~/>P/r,~\io/ I-‘.\!(h,,/ 93: 820-830, 1979. 5. Dixon, W. J. and M. B. Brown tedrtors). Birwrdic trl C’rwrprrc,r Progrom~. P-Series. Berkeley. (‘A: University of California Press, 1979.

MEAL PATTERN

ANALYSIS

6. Jacobs, H. L. and K. L. Sharma. Taste versus calories: sensory and metabolic signals in the control of food intake. Ann NY Acad Sci 1sI: 10861125, 1%9. 7. Le Magnen, J. and S. Tallon. La periodicite spontanee de la prise d’aliments ad libitum du rat blanc. J Physiol (Paris) 5% 323-349, 1966. 8. Le Magnen, J. and M. Devos. Metabolic correlates of the meal onset in the free food intake of rats. Physiol Behav 5: 805-814, 1970. 9. Le Magnen, J. and M. Devos. Meal to meal balance in rats. Physiol Behav 32: 39-44, 1984. 10. Levitsky, D. Feeding conditions and intermeal relationships. Physiol Behav 5: 291-309, 1970.

443

11. Mayer, J. General characteristics of the regulation of food intake. In: Handbook of Physiology, Sec. 6, Alimentary Canal, Vol I, edited by C. F. Code. Washington DC: American Physiological Society, 1967, pp. 3-9. 12. Panksepp, J. Reanalysis of feeding patterns in the rat. J Comp Physiol Psycho1 Sz: 78-91, 1973. 13. Simpson, S. J. and E. A. Bemays. The regulation of feeding: Locusts and blowflies are not so different from mammals. Appetite: 4: 313-346, 1983. 14. Slater, P. J. B. and N. P. Lester. Minimising errors in splitting behaviour into bouts. Behaviour 79: 153-162, 1982. 15. Thomas, D. W. and J. Mayer. Meal taking and regulation of food intake by normal and hypothalamic hyperphagic rats. J Comp Physiol Psycho1 66: 642-653, 1968.