Total and Percentage REM Sleep Correlate with Body Weight in 36 Middle-Aged People

Transcription

1 Sleep 10(1):69-77, Raven Press, New York 1987, Association of Professional Sleep Societies Total and Percentage REM Sleep Correlate with Body Weight in 36 Middle-Aged People Kirstine Adam University Department of Psychiatry, Royal Edinburgh Hospital, Edinburgh, Scotland Summary: Thirty-six volunteers of mean age 52 years had their sleep recorded in the laboratory on 5 consecutive nights, Eight factors were investigated as possible correlates of percentage REM sleep and of total minutes of REM sleep: age, height, weight, bedtime, arising time, oral temperature at bedtime and on arising, and total sleep time, REM percentage correlated significantly with body weight and with oral temperature at bedtime and correlated negatively with age, Owing to intercorrelations among the factors investigated, partial correlational analyses were done and revealed that body weight was the only factor that remained significantly correlated with REM percentage, Similarly, total sleep time and body weight were the only factors that remained significantly correlated with total minutes of REM sleep, whereas age and bedtime temperature just missed significance when other factors were held constant. Key Words: REM sleep-body weight. Despite intense research interest we do not know what purpose REM sleep serves, A role in brain metabolism is suggested by the very large amounts of REM sleep that babies have when their brains are maturing rapidly (1), the prolonged increase in the amount of REM sleep during recovery from poisoning by ens drugs (2), and the reduced amounts of REM sleep recorded in elderly chronic brain syndrome patients as compared with age-matched controls (3). The last study suggests that cognitive ability is a more important correlate of REM duration than age, a conclusion supported by Prinz (4) but not by Spiegel (5), nor by our own study of 47 normal children, in whom no relationship between REM sleep and intelligence was found (6). Research studies such as those above and theoretical discussions about the function of REM sleep (7-11) have concentrated on the brain, largely neglecting the possibility of somatic functions for REM sleep, though Horne (12) has suggested a role for REM sleep in energy conservation. Among a group of 16 healthy middle-aged volunteers; I earlier found a significant correlation between body weight and REM sleep averaged over 20 nights for each person (13). This was true of the total minutes of REM sleep and the percentage of total sleep time spent in REM sleep. In the same group there was no significant correlation Accepted for publication July Address correspondence and reprint requests to Dr. K. Adam at University Dept. Psychiatry, Royal Edinburgh Hospital, Morningside Park, Edinburgh EHIO 5HF, Scotland. 69

2 70 K. ADAM between intelligence quotient (range ) and REM duration or REM percentage (14). Spiegel (15) found no significant correlation between amounts of REM sleep and body weight, and Ohkawa and Nakazawa (16) reported a negative correlation between these variables. However, I believe that both these studies have methodological flaws that may have accounted for the results (see Discussion). The present study investigates further the relationship between REM sleep and body weight, this time taking into account other factors that might influence the amount or percentage of REM sleep, such as age, customary times of going to bed and getting up, total sleep time, and body temperature. The last of these is included because not only does the sleep/wake cycle influence both the amplitude and the phasing of the circadian body temperature rhythm (17), but also the body temperature rhythm may exert an effect on sleep onset latency, sleep duration (18), and timing of REM sleep (19). METHODS Subjects and data collection Thirty-six volunteers (26 women and 10 men) aged years (mean 52 years) slept in comfortable, temperature-regulated bedrooms in the sleep laboratory on 5 consecutive nights. None of these subjects had been involved in the previous study of REM sleep and body weight, none had taken CNS drugs in the preceding months, and none took daytime naps. Subjects went to bed and got up at their usual times and spent an average of 7 h 55 min in bed. Oral temperatures were taken (5 min sublingually) just before a subject's selected time for lights off and immediately after lights on in the morning on each of the 5 study nights. Eating, drinking, bathing, smoking, and teeth cleaning were avoided for 30 min prior to taking an oral temperature. The electroencephalogram, electro-oculogram (EOG), and electromyogram were recorded from lights out to lights on each night. Each subject was weighed and had his/her height measured on one of the mornings. Twenty-three of the subjects were within 10% of their ideal body weight for their height, eight were within 10-20%, four were between 20 and 30% overweight, and one was 38.6% overweight according to the Metropolitan Life Insurance Co. (20). The 180 overnight recordings of sleep were scored blind according to the criteria of Rechtschaffen and Kales (21). Recordings were checked by an independent observer for electroencephalographic activity indicative of hypnotic drug intake, and all patients were judged to be drug-free. All 5 consecutive nights for each subject were used in the analysis, as it was reasoned that amounts of REM sleep are, at least partly, homeostatically regulated such that a low amount of REM sleep on one night may be compensated for over the next night or two by above average amounts of REM sleep for that individual. Statistical analysis The BMDP (BioMeDical Programs) 1983 P-series computer programs of the Health Sciences Computing Facility (UCLA, Los Angeles, CA, U.S.A.) were used to analyze the data. First, Pearson product-moment correlations were calculated among the variables investigated, i.e., body weight, age, height, time on the clock for going to bed (bedtime), bedtime oral temperature, time to arise, oral temperature on arising in the morning, total sleep time, and total and percentage REM sleep. Second, because of the intercorrelations among these variables, partial correlation coefficients were calculated Sleep, Vol. 10, No.1, 1987

3 CORRELATES OF AMOUNT OF REM SLEEP 71 for mean percentage REM and for mean total REM sleep with each of the other individual factors in turn (i.e., holding the other seven factors constant). RESULTS Summary of statistics and correlations Table 1 summarizes the means, standard deviations, and range of values of the variables used in the correlation analyses. Table 2 lists the Pearson product-moment correlation coefficients found between pairs of variables. The mean percentage of total sleep time spent in REM sleep was significantly correlated with the following: body weight, mean oral temperature at bedtime, and age (negatively). The correlation with height approached statistical significance. The mean total minutes of REM sleep (Table 2) correlated significantly with mean total sleep time, mean oral temperature at bedtime, body weight and negatively with mean time for lights out, and age. Some of the factors that were significantly correlated with the two measures of REM sleep were also significantly correlated with each other (Table 2). For example, in this group of subjects, body weight was negatively correlated with age. Similarly, mean total sleep time was negatively correlated with the average time on the clock for going to bed, and mean oral temperature at bedtime was negatively correlated with the average time for going to bed. There was a significant negative correlation between age and height in these subjects. Because of these intercorrelations among the "independent" variables, partial correlation coefficients were then calculated for mean percentage REM sleep (Table 3) and mean total minutes REM sleep (Table 4). Partial correlations Table 3, when compared with Table 2, shows that the statistical significance of the coefficient of correlation between percentage REM sleep and body weight was virtually unchanged when all the other factors were partialed out, whereas the correlation between percentage REM sleep and age became insignificant when the other factors were partialed out. The correlation between percentage REM sleep and oral tempera- TABLE 1. Means, standard deviations, and range of values over 36 subjects of the variables used in the correlation analyses Variable Percentage REM sleepa (%) Total REM sleepa (min) Age (years) Height (mm) Weight (kg) Oral temperature at bedtime a cae) Oral temperature on arising a cac) Bedtime a Arise time a Total sleep time a (min) a Mean of 5 consecutive nights. Mean ± SD 21.4 ± ± ± 7 1,664.5 ± ± ± ± h ± 37 min 0705 h ± 34 min ± 31.7 Range of values ,530-1, h h Sleep, Vol. 10, No

4 72 K. ADAM TABLE 2. Pearson's product-moment correlation coefficients (r) between pairs of variables Total REM Age Height Weight REM % 0.90" -0.44" b Total REM b ' Age a b Height 0.69 a Weight Bedtime oral temperature Morning oral temperature Total sleep time Bedtime a Significant at p < level, df = 34. two tailed. b Significant at p < 0.01 level, df = 34, two tailed., Significant at p < 0.05 level. df = 34, two tailed. Bedtime oral temperature 0.40' 0.39' Morning Total oral sleep Arise temperature time Bedtime time b -0.35' ' ' b ture at bedtime was lower after partial correlational analysis and just failed to reach the 0.05 criterion of significance. Table 4, when compared with Table 2, shows that the significant correlation between total minutes of REM sleep and total sleep time remained virtually unchanged when the other factors were partialed out. The correlation between total minutes of REM sleep and body weight remained significant after partialing out the effects of all the other factors investigated. The correlation between age and total minutes of REM sleep just failed to reach the 0.05 criterion when the effects of the other factors were partialed out. Similarly, the partial correlation coefficient between oral temperature at bedtime and total minutes of REM sleep just failed to reach the 0.05 criterion. Post hoc analyses Correlations on each of the 5 nights. The Pearson correlation coefficients between REM percentage and body weight were significant for 4 of the 5 consecutive nights (df = 34; r = 0.36,0.47, 0.37, 0.12, 0.39), whereas between, e.g., bedtime temperatures and REM percentage in the sleep that followed, no correlation coefficient reached significance (df = 34; r = 0.28, 0.29, 0.18, 0.25, 0.10). Male/female data treated separately. In the group of 26 women, percentage REM sleep (mean 20.9%, range %) was significantly correlated (r = 0.44, df = 24, TABLE 3. Partial correlation coefficients between mean percentage REM sleep and each of the eight variables in turn, holding the other seven constant Partial correlation coefficient with REM %, dj = 33 Age Height Weight Bedtime temperature Morning temperature Total sleep time Bedtime Arise time , P < Sleep, Vol. 10, No. 1,1987

5 CORRELATES OF AMOUNT OF REM SLEEP 73 TABLE 4. Partial correlation coefficients between mean total minutes REM sleep and each of the eight variables in turn, holding the other seven constant Partial correlation coefficient with total minutes REM, df = 33 Age Height Weight Bedtime temperature Morning temperature Total sleep time Bedtime Arise time , P < , P < p < 0.03) with body weight (mean 60.6 kg, range kg). Similarly, total minutes of REM sleep was significantly correlated with body weight (r = 0.41, p < 0.05). Within the group of 10 men (mean body weight 75.0 kg, range kg), the same trends were found but, perhaps owing to the small number of subjects, were not significant, i.e., REM percentage (mean 22.6%, range %) with body weight (r = 0.31, df = 8, NS). Total minutes of REM sleep was also positively correlated with body weight (r = 0.25, df = 8, NS). Data from present and previous studies combined. Figure 1 shows the mean percentage REM sleep versus body weight data for the 36 subjects in the present study and for 16 subjects reported in an earlier study (13). The correlation between body weight and percentage REM sleep for the combined data (i.e., 52 subjects) was also significant (r = 0.44, df = 50, p < ). The mean ratio of percentage REM sleep to kilograms of body weight was 0.33 ± 0.05 in the present study and 0.29 ± 0.04 in the previous study. DISCUSSION The significant positive correlation between body weight and REM sleep in middleaged volunteers previously reported (13) has been confirmed with different subjects and with a different design. In the previous study all 16 subjects were recorded for 8.75 h on each of 20 nights, whereas in the present study the lengths of recordings varied as they were determined by subjects' customary times for going to bed and getting up. Despite these differences in design, the results are similar. Simple product-moment correlations at first suggested that age was an important correlate of the amount of REM sleep (both as a duration and as a percentage); but when the other factors were partialed out, the correlations between age and measures of REM sleep just failed to reach the 0.05 criterion of significance. It is not possible to eliminate the effects of aging on REM time and REM percentage in the present study because of the intercorrelation between age and weight, though it can be said that body weight had the greater influence: Age became nonsignificant in the partial correlations. Other studies suggest that across age groups from the twenties to sixties REM percentage is relatively stable, only falling slightly in advanced age (22). In the previous study in which a significant correlation between body weight and Sleep, Vol, 10, No, I, 1987

6 74 K.ADAM T : Present study r=0'~5, (.)df=3~,p<o 01 df = 1~, p<o OI 25 (0) Previous study r=0'63, a.., 0 'iii :::;: w Cl::., 20 C., 01 C ~., a. 0 0 c., :::;: o Body weight ( Kg) FIG. 1. Correlation between mean percentage REM sleep and body weight in the 36 subjects of the present study (percentage REM averaged over 5 nights) compared with the same correlation in a previous study of 16 subjects (13) (REM sleep averaged over 20 nights). REM sleep was reported (13), the age range was narrower than in the present study (15 vs. 25 years), and in the former study no significant correlation betweeen age and total or percentage REM sleep was found (14). The consistency of the relationship between body weight and REM sleep is demonstrated by the fact that the correlation coefficients between the measures were statistically significant on 4 of the 5 nights recorded. The men in this study were heavier and had, on average, more REM sleep, but the correlations between body weight and REM sleep in the group as a whole were not attributable just to differences between the sexes, for the significant relationships between body weight and measures of REM sleep remained when the data from the 26 women were considered separately. Oral temperature at bedtime was the other factor that was significantly correlated with total and percentage REM sleep. However, when the effects of the other factors were partialed out, the values of the correlation coefficients fell to just below the 0.05 level of significance. Partialing out the effect of time of going to bed, which was significantly correlated with oral temperature at bedtime, may have contributed to this. As mentioned in the introductory section, there have been studies in which a significant positive correlation between amounts of REM sleep and body weight was not found. Before I discuss these studies, it is worth listing certain experimental requirements that I believe to be necessary before amounts of REM sleep can be meaningfully correlated with body weights. Some of these requirements may seem obvious; however, neither of the studies referred to fulfill all of them: There should be a large number of subjects, with a wide range of body weights. The sleep data should be reli- Sleep, Vol. 10, No. I, 1987

7 CORRELATES OF AMOUNT OF REM SLEEP 75 able, i.e., the average of several nights, preferably consecutive. The sleep data should be as free from extraneous influences as possible; e.g., subjects should not have taken ens drugs for at least 6 weeks (23) and data collected on nights on which blood samples were taken via an intravenous catheter should not be used (24). The weakness of my previous study of REM sleep and body weight was the relatively small number of subjects (n = 16). However, this was mitigated to some extent by the reliability of the sleep data (an average of 20 nights per person recorded as four blocks of 5 consecutive nights). Even the 36 subjects of the present study constitute a relatively small number for a correlation analysis, but there was a wide distribution of body weights and the sleep data were based on the average of 5 consecutive nights. The findings of the present and previous studies (13) contrast with those of Spiegel (15). He attempted to explain his contrary results in terms of "strikingly low REM durations" in the female subjects of my earlier study (13). Apart from one subject (see Fig. I) whose mean total REM duration was 49.2 min, the women had normal amounts of REM sleep (range min) as did the men ( min). Spiegel's study fulfilled at least two of the criteria: He had 57 subjects and recorded 4 consecutive nights. However, he advised his subjects "not to omit any customary naps" and, much more importantly, 16 of his subjects had recently taken or were still taking centrally acting drugs during the week when the recordings were made. This means that the REM sleep values from these subjects were probably altered from normal, thus distorting the data used to calculate the correlation coefficient. He did not present the range of body weights. Ohkawa and Nakazawa (16) reported a significant negative correlation between percentage REM sleep averaged over 2 nights and body weight among 11 young men. However, inspection of the figure in their article shows that 7 of the 11 young men had very similar body weights and percentages of REM sleep and suggests that the correlation coefficient reported was greatly affected by outliers; e.g., one subject was very much heavier than the rest (84 kg), but happened to have the second lowest percentage REM sleep. If REM sleep and body weight are positively correlated, why should this be? Bearing in mind that biological systems have multiple influences, a possible explanation could be that a large proportion of body weight is due to skeletal muscle mass, and, as a characteristic feature of REM sleep is loss of muscle tone, it could be speculated that the relation of body weight and REM sleep suggests a compensation for the high rate of oxidative metabolism in muscle during daytime activity. The fact that, at least in the short term, increased physical activity does not lead to more REM sleep (25) seems to rule out this idea. However, the studies reviewed (25) looked only at the acute effects of exercise, and it may be that an increase in slow wave sleep takes priority over an increase in REM sleep in the short term. Others have seen sleep in general (26,27) or REM sleep in particular (12) as a time of energy conservation. The significant positive correlations found (28) between total sleep time and metabolic rate across species (oxygen consumed per kilogram of body weight) support the view that there is a relationship between the rate of energy expenditure and the amount of sleep taken per 24 h, thereby implying a negative correlation between total sleep time and gross body weight across species, as there is a negative correlation between metabolic rate per kilogram and body weight across species (29). In the energy conservation model for REM sleep (12), a negative correlation across species between body weight and REM percentage was suggested, and so the positive Sleep, Vol. 10, No.1, 1987

8 76 K.ADAM correlation between body weight and REM percentage that I have reported does not appear to conform to this hypothesized role for REM sleep. However, it may be that within a species, the controlling factors in the relationship between REM percentage and body weight are different from those operating across species. It is known that in humans, weight gain is associated with longer sleep and weight loss with shorter sleep (30), i.e., the opposite of what would be expected from the across-species correlation (28). It may be that within a species, the positive correlation between total daily energy expenditure and gross body weight (29) has a greater influence than the negative correlation between metabolic rate per kilogram and body weight that seems to be so important across species. Thus, REM sleep could have a role associated with whole-body energy balance, or perhaps there is some other factor upon which both REM sleep and body weight are mutually related. At a practical level, if amounts of REM sleep and body weights are correlated, it means that in studies where the sleep of one group is compared with that of another (e.g., good versus poor sleepers), the body weights of the two groups should be matched, otherwise spurious differences in their sleep may be found. Acknowledgment: Thanks are owed to the Scottish Hospital Endowments Research Trust, Schering AG of Berlin, Prof. Ian Oswald, and Maureen Tomeny. REFERENCES I. Roffwarg HP, Muzio ln, Dement WC. Ontogenetic development of the human sleep-dream cycle. Science 1966;152: Haider 1, Oswald I. Late brain recovery processes after drug overdose. Br Med J 1970;2: Feinberg 1, Koresko RL, Heller N. EEG sleep patterns as a function of normal and pathological aging in man. J Psychiatr Res 1967;5: Prinz P. Sleep patterns in the healthy aged: relationship with intellectual function. J Gerontal 1977;322: Spiegel R. Aspects of sleep, daytime vigilance, mental performance and psychotropic drug treatment in the elderly. Gerontology 1982;28: Borrow Sl, Adam K, Chapman K, Oswald I, Hudson L, Idzikowski Cl. REM sleep and normal intelligence. Bioi Psychiatry 1980;15: Oswald 1. Human brain protein, drugs and dreams. Nature 1969;223: Oswald 1. Sleep, the great restorer. New Scientist 1970;46: Hartmann E. The functions of sleep. New Haven: Yale University Press, Stern WC, Morgane Pl. Theoretical view of REM sleep function: maintenance of catecholamine systems in the central nervous system. Behav Bioi 1974; 11: McGinty 01, Drucker-Colin RR. Sleep mechanisms: biology and control of REM sleep. Int Rev Neurobioi 1982;23: Horne la. Factors relating to energy conservation during sleep in mammals. Physiol Psychol 1977;5: Adam K. Body weight correlates with REM sleep. Br Med J 1977;1: Adam K. No significant correlation between 1.Q. and the amount of REM sleep. PhD thesis, University of Edinburgh, 1977, pp Spiegel R. Sleep and sleeplessness in advanced age. New York: Spectrum, 1981: (Advances in sleep research; vol 5). 16. Ohkawa T, Nakazawa Y. Correlations of some physical variables with REM sleep and slow wave sleep in man. Folia Psychiatr Neurol Jpn 1982;36: Mills ln, Minors OS, Waterhouse 1M. The effect of sleep upon human circadian rhythms. Chronobiologia 1978;5: Zulley 1, Wever R, Aschoff 1. The dependence of onset and duration of sleep on the circadian rhythm of rectal temperature. Pj1ugers Arch 1981 ;391 : Czeisler CA, Zimmerman lc, Ronda 1M, Moore-Ede MC, Weitzman ED. Timing of REM sleep is coupled to the circadian rhythm of body temperature in man. Sleep 1980;3: Metropolitan Life Insurance Co., New York. Mortality among overweight men and women. Stat Bull 1960;41:(Feb) 6, (Mar) I. Sleep, Vol. 10. No.1, 1987

9 CORRELATES OF AMOUNT OF REM SLEEP Rechtschaffen A, Kales A, eds. A manual of standardized terminology, techniques and scoring system for sleep stages of human subjects. Washington, DC: US Government Printing Office, Williams RL, Karacan I, Hursch CJ. Electrocephalography (EEG) of human sleep: clinical applications. New York: John Wiley & Sons, Oswald I, Priest RG. Five weeks to escape the sleeping-pill habit. Br Med J 1965;2: Adam K. Sleep is changed by blood sampling through an indwelling venous catheter. Sleep 1982;5: Torsvall L. Sleep after exercise: a literature review. J Sports Med 1981 ;21 : Snyder F. Toward an evolutionary theory of dreaming. Am J Psychiat 1966;123: Berger RJ. Bioenergetic functions of sleep and activity rhythms and their possible relevance to ageing. Fed Proc 1975;34: Zepelin H, Rechtschaffen A. Mammalian sleep, longevity and energy metabolism. Brain Behav Evol 1974;10: Kleiber M. The fire of life. An introduction to animal energetics. Huntington, NY: Robert E. Kreiger, 1975: Crisp AH, Stonehill E. Aspects of the relationship between psychiatric status, sleep, nocturnal motility and nutrition. J Psychosom Res 1971 ;15: Sleep, Vol. 10, No. I, 1987