Sleep and eating behavior in adults at risk for type 2 diabetes

Abstract

Insufficient quantity and quality of sleep may modulate eating behavior, everyday physical activity, overall energy balance, and individual risk of obesity and type 2 diabetes. We examined the association of habitual sleep quantity and quality with the self-reported pattern of eating behavior in 53 healthy urban adults with parental history of type 2 diabetes (30F/23M; mean [SD] age: 27 [4] y; BMI: 23.9 [2.3] kg/m²) while taking into consideration the amount of their everyday physical activity. Participants completed 13 [3] days of sleep and physical activity monitoring by wrist actigraphy and waist accelerometry while following their usual lifestyle at home. Overnight laboratory polysomnography was used to screen for sleep disorders. Subjective sleep quality was measured with the Pittsburgh Sleep Quality Index. Eating behavior was assessed using the original 51-item and the revised 18-item version of the Three Factor Eating Questionnaire including measures of cognitive restraint, disinhibition, hunger, and uncontrolled and emotional eating. In multivariable regression analyses adjusted for age, BMI, gender, race/ethnicity, level of education, habitual sleep time measured by wrist actigraphy and physical activity measured by waist accerelometry, lower subjective sleep quality was associated with increased hunger, more disinhibited, uncontrolled and emotional eating, and higher cognitive restraint. There was no significant association between the amount of sleep measured by wrist actigraphy and any of these eating behavior factors. Our findings indicate that small decrements in self-reported sleep quality can be a sensitive indicator for the presence of potentially problematic eating patterns in healthy urban adults with familial risk for type 2 diabetes.

INTRODUCTION

The rise in obesity-related morbidity in modern society reflects a number of environmental and behavioral changes that facilitate overeating and physical inactivity. Several aspects of eating behavior, including patterns of food intake driven by hunger, cognitive restraint, loss of voluntary control and emotional distress, are thought to modify individual propensity to overeat (1–3). Better understanding of the relationship between these eating patterns and other obesity-promoting behaviors may facilitate the development of improved strategies for successful weight maintenance and metabolic risk reduction.

Sleep is an important health-related factor that may influence human eating behavior (4). Today, many Americans sleep less than 6 h per night and some (5), but not other (6, 7), epidemiologic studies have associated such short sleep duration with increased incidence of obesity. Studies of healthy volunteers in the laboratory indicate that short-term sleep restriction can modify energy intake and expenditure, fuel metabolism, and concentrations of circulating hormones that affect hunger and appetite (8–16). However, the contribution of such acute experimentally-induced changes in hunger and energy metabolism to the association between self-reported short sleep and obesity in free-living adults is poorly understood (4, 17). There is also concern that the relationship between self-reported short sleep and obesity in epidemiologic studies may be confounded by factors, such as undiagnosed sleep problems (e.g. sleep apnea, insomnia) (18, 19), poor physical and emotional health (18, 19), and systemic bias in the subjective recall of sleep and physical activity (20, 21).

Adults with parental history of type 2 diabetes have a high risk for developing the disease, which is exacerbated by physical inactivity and excessive weight gain (22). Prevention of obesity in the offspring of diabetic patients is accompanied by a reduction in excess risk of type 2 diabetes by nearly 40% (22). Thus, it is important to understand the relationship between food intake, physical activity, and sleep in this high-risk population (23). Recent data from our laboratory suggest that urban adults with parental history of type 2 diabetes who habitually curtail their sleep have reduced everyday physical activity (24). Whether habitual sleep patterns could be related to potentially problematic patterns of eating behavior in this high-risk population is not known. Therefore, we examined the association of habitual sleep quantity and quality with patterns of eating behavior characterized by cognitive restraint, hunger, loss of voluntary control, and emotional eating in urban adults with parental history of type 2 diabetes, while taking into consideration the amount of their everyday physical activity.

METHODS AND PROCEDURES

Participants

Healthy men and women between the ages of 21 and 40 y with body mass index between 19 and 27 kg/m², who lived in the greater Chicago area and had at least one parent with type 2 diabetes, were recruited through local advertisements. Volunteers who passed a brief telephone interview were invited for screening in our Clinical Research Center. Body weight and height were measured after an overnight fast using a calibrated medical scale (Scale-Tronix, Wheaton, IL) and a stationary Harpenden stadiometer (Holtain, Crymych, Wales) with participants dressed in light clothing without shoes. Individuals were excluded from participation if they had abnormal findings on medical history, physical examination and routine screening tests (complete blood counts, comprehensive metabolic and thyroid function panels, 12-lead ECG, 75-g oral glucose tolerance test); depressed mood (Center for Epidemiologic Studies of Depression, CES-D (25), score > 15 confirmed by clinical interview); pregnancy or childbirth during the last year; night-shift work, frequent travel across time zones, or self-reported sleep problems (Pittsburgh Sleep Quality Index, PSQI (26), global score > 5); use of tobacco, excess alcohol (>14 drinks/week for men; >7 for women), or prescription, over-the-counter and illegal drugs and supplements that can effect sleep and eating behavior. Research volunteers gave written informed consent and were paid for their participation.

Study protocol

The study protocol was approved by the Institutional Review Board of the University of Chicago. Enrolled participants were asked to complete 14 days of sleep monitoring while following their usual lifestyle at home. A small accelerometer equipped with an event marker (Actiwatch-64, Mini-Mitter Respironics, Bend, OR) was attached to a wrist band on their non-dominant arm and actigraphy data were collected continuously in 1-minute epochs to measure sleep duration under free-living conditions (27). Since some prior epidemiologic studies have found an association between self-reported sleep and physical activity (28–31), we also measured the amount of body movement of each participant using a small waist accelerometer (Actical, Mini-Mitter-Respironics, Bend, OR). Physical activity data from a subset of the participants in this study have been reported elsewhere (24).

After the 2-week home monitoring period, participants were scheduled to complete one night of laboratory polysomnography (Neurofax-1100 EEG Acquisition System, Nihon-Kohden) including electroencephalography, electrooculography, electromyography, airflow, thoracic and abdominal respiratory effort, electrocardiography and pulse oximetry to exclude the presence of primary sleep pathology, sleep movement disorder, or sleep disordered breathing (respiratory disturbance index > 10 or sleep apnea index > 3). Sleep was scheduled between 23:00–24:00 and 7:30–8:30 with a fixed time-in-bed of 8.5 h. Records were scored in 30-second epochs of wake, movement, stage 1, 2, 3, 4, and rapid-eye-movement sleep according to standard criteria. Respiratory events, periodic leg movements, and arousals were scored using current clinical guidelines.

Data analysis and statistics

Home activity records were analyzed using version 2.12 of the software provided with the Actical device. The total number of activity counts during each 24-h period was averaged across all recorded days to obtain a measure of individual physical activity. Nighttime sleep was scored automatically with Actiware Sleep version 3.4 provided with the Actiwatch using a medium sensitivity setting of 40. Habitual sleep duration was calculated as the average number of minutes scored as sleep across all recorded nights. Subjects with less than 6 nights of data were not included in the analysis. The average sleep fragmentation index of each individual wrist actigraphy data set was used as a measure of habitual sleep quality. The index is calculated during the nighttime sleep period by adding the percentage of time spent in non-sleep epochs containing above-sleep-threshold amounts of movement and the percentage of time spent in brief periods with sub-sleep-threshold wrist movement that last only 1 min.

The sum of all epochs scored as sleep was used to measure the amount of overnight sleep in the laboratory. Measures of laboratory sleep quality included the number of arousals per hour of sleep (arousal index), the number of awakenings and the amount of wake time during the night. The Pittsburgh Sleep Quality Index (PSQI) score of global sleep disturbance was used as a measure of self-reported sleep quality (26). Higher scores on this scale reflect lower subjective sleep quality.

During the initial screening visit, all participants completed the 51-item Three-Factor Eating Questionnaire (TFEQ), which measures three dimensions of eating behavior: cognitive restraint of eating, disinhibition, and hunger (32). Karlsson et al. have revised and abbreviated the original 51-item TFEQ to improve its scaling properties and construct validity and we used their 18-item TFEQ version (33) to derive additional scores of cognitive restraint (tendency to consciously restrict food intake to control body weight), uncontrolled eating (tendency to eat more because of loss of control over food intake), and emotional eating (tendency to overeat related to dysphoric mood) (34).

All statistical analyses were performed using SPSS version 18.0 (SPSS Inc., Chicago, IL). Multivariable linear regression models adjusted for age, gender, BMI, race/ethnicity and level of education (as a surrogate of socioeconomic status) were used to examine the relationship between each eating behavior factor as a dependent variable and each of 3 predictor variables: habitual sleep duration (measured by wrist actigraphy), subjective sleep quality (PSQI score), and free-living physical activity (measured by waist accelerometry). Since self-reported sleep quality and physical activity emerged as significant predictors of several aspects of eating behavior, the role of each of these 3 predictor variables was re-examined after control for the other two was added to the initial regression models. Finally, partial correlation analysis, controlling for age, gender, race/ethnicity, BMI, level of education, and objectively-measured sleep duration and physical activity, was used to explore the relationship of subjective sleep quality (PSQI score) with actigraphy- and polysomnography-based measures of sleep structure and quality and self-ratings of depressed mood (CES-D score).

RESULTS

Fifty three participants completed an average of 13 (SD 3) days of home sleep monitoring. Participant characteristics are summarized in Table 1. The average sleep time of the participants measured by home actigraphy ranged between 4 h 33 min and 8 h 14 min per night. Thirty eight percent of the participants habitually slept < 6 h/night. All subjects had good subjective sleep quality with PSQI scores ranging between 0 and 5 (26). Overnight laboratory polysomnography in 48 (91%) of the participants who kept their study appointments showed no sleep pathology. The sleep architecture of the participants was typical for healthy individuals monitored by full polysomnography under laboratory conditions without prior habituation (Table 1).

Table 1.

Participant characteristics and measures of eating behavior, sleep, and activity

Participant characteristics
Number of participants	53 (30F/23M)
Caucasian/African American/Asian/Hispanic	29/13/7/4
Age (y)	27 (4)
BMI (kg/m2)	23.9 (2.3)
Level of education (y)	17 (2)
Depressed mood (CES-D score)	4 (2; 8)
Eating behavior factors
Cognitive restraint score (51-item TFEQ)	9 (5; 12.5)
Disinhibition score (51-item TFEQ)	4 (3; 6)
Hunger score (51-item TFEQ)	3 (1.5; 5)
Cognitive restraint score (18-item TFEQ)	2 (1; 4)
Uncontrolled eating score (18-item TFEQ)	1 (0; 2)
Emotional eating score (18-item TFEQ)	0 (0; 2)
Subjective sleep quality
Global sleep disturbance (PSQI score)	2 (1; 3)
Free-living sleep and activity monitoring
Measured sleep duration (min/day)	379 (53)
Sleep fragmentation index (%)	31 (10)
Total activity counts (thousands/day)	217 (117)
Laboratory polysomnography
Number of participants	48 (27F/21M)
Sleep onset latency (min)	32 (30)
Total sleep time (min)	439 (48)
Stage 1 sleep (min)	30 (17)
Stage 2 sleep (min)	262 (41)
Slow wave sleep (stages 3+4, min)	51 (33)
Rapid-eye-movement sleep (min)	100 (32)
Wake after sleep onset (min)	43 (27)
Sleep efficiency (%)	87 (8)
Arousal index (events/hour)	13 (7)
Number of awakenings	4 (3)
Respiratory disturbance index (events/hour)	3 (4)

Data are reported as mean (SD) for continuous variables and median (inter-quartile range) for questionnaire based scores. BMI: body mass index; CES-D: Center for Epidemiologic Studies of Depression scale; TFEQ: Three-Factor Eating Questionnaire; PSQI: Pittsburgh Sleep Quality Index.

Reduced subjective sleep quality (higher PSQI score) was associated with eating behaviors characterized by increased hunger, uncontrolled and emotional eating, and more cognitive restraint (Table 2; Model 1). In contrast, there was no significant association between sleep duration measured by wrist actigraphy and any of these eating behavior factors (Table 2). Higher levels of habitual physical activity were associated with a pattern of more uncontrolled and hunger-dependent eating (Table 2). The association of reduced subjective sleep quality (higher PSQI score) with eating behaviors characterized by increased hunger and cognitive restraint, and uncontrolled and emotional eating remained qualitatively and quantitatively similar when regression analyses controlled for habitual sleep duration measured by wrist actigraphy and physical activity measured by waist accelerometry (Table 2; Model 2).

Table 2.

Subjective sleep quality and measured sleep duration and physical activity as predictors of eating behavior

	Model 1			Model 2
	Sleep qualitya	Sleep durationb	Physical activityc	Sleep qualitya	Sleep durationb	Physical activityc
Restraint (TFEQ-51)
B	3.1	−1.0	−0.3	3.5	−1.1	0.0
95% CI	−0.1 to 6.3	−2.8 to 0.7	−1.5 to 1.0	0.3 to 6.7	−2.8 to 0.6	−1.3 to 1.2
R2 change	0.067	0.026	0.004	0.093	0.032	0.000
P	0.053	0.235	0.675	0.032	0.198	0.948
Disinhibition
B	1.9	−0.4	0.0	1.9	−0.5	0.1
95% CI	0.5 to 3.4	−1.2 to 0.4	−0.6 to 0.6	0.5 to 3.3	−1.3 to 0.3	−0.4 to 0.7
R2 change	0.119	0.017	0.000	0.122	0.031	0.003
P	0.010	0.341	0.949	0.011	0.190	0.658
Hunger
B	1.9	0.5	1.0	2.1	0.1	1.0
95% CI	0.2 to 3.7	−0.4 to 1.5	0.3 to 1.7	0.4 to 3.7	−0.8 to 1.0	0.4 to 1.6
R2 change	0.087	0.024	0.169	0.097	0.001	0.158
P	0.034	0.278	0.004	0.018	0.776	0.003
Restraint (TFEQ-18)
B	1.3	−0.1	−0.3	1.4	0.0	−0.3
95% CI	0.1 to 2.5	−0.8 to 0.5	−0.8 to 0.2	0.2 to 2.6	−0.7 to 0.6	−0.7 to 0.2
R2 change	0.088	0.002	0.037	0.104	0.000	0.031
P	0.029	0.745	0.184	0.023	0.921	0.204
Uncontrolled eating
B	1.4	0.2	0.5	1.6	0.0	0.5
95% CI	0.3 to 2.5	−0.4 to 0.9	0.1 to 0.9	0.5 to 2.7	−0.6 to 0.6	0.1 to 0.9
R2 change	0.112	0.012	0.101	0.151	0.000	0.103
P	0.017	0.451	0.030	0.005	0.921	0.019
Emotional eating
B	0.8	−0.1	0.0	0.7	−0.1	0.0
95% CI	0.2 to 1.4	−0.4 to 0.3	−0.2 to 0.2	0.1 to 1.3	−0.4 to 0.2	−0.2 to 0.3
R2 change	0.119	0.002	0.001	0.100	0.008	0.003
P	0.010	0.761	0.846	0.025	0.501	0.682

Model 1 was adjusted for age, gender, race/ethnicity, BMI and years of education. Model 2 controlled for the remaining 2 predictors in addition to the variables included in Model 1. B - regression coefficient reflecting the change in the dependent variable for:

^a1-point increase in the square-root transformed PSQI score;

^b1-h increase in measured sleep duration;

^c100,000-count increase in average daily body movement (bold numbers show significant associations). 95% CI - 95% confidence interval for B.

Partial correlation analysis to explore the correlates of self-rated sleep quality showed that higher PSQI scores were associated only with more depressed mood (CES-D score; R=0.33; P=0.036). Using CES-D instead of PSQI as a predictor in our fully-adjusted regression analysis (see Model 2 in Table 2) showed that the score for depressed mood was associated with hunger (B: 1.0; 95%CI: 0.2 to 1.7; P=0.014) and uncontrolled eating (B: 0.6; 95%CI: 0.1 to 1.1; P=0.020). The association of subjective sleep quality with restrained, disinhibited, uncontrolled, and emotional eating did not change qualitatively or quantitatively when control for CES-D was included in Model 2, whereas the association with hunger was attenuated (B: 1.5; 95%CI −0.2 to 3.2; P=0.084). There was no significant relationship between subjective sleep quality and wrist actigraphy- or laboratory polysomnography-based measures of sleep quality (sleep fragmentation index, arousal index, number of awakenings, and wake time during the night) or other sleep architecture indices.

DISCUSSION

This study examined the relationship of habitual duration and quality of sleep with several dimensions of eating behavior in free-living urban adults with parental history of type 2 diabetes. Our results show that decrements in self-reported sleep quality are associated with eating patterns characterized by increased hunger, uncontrolled and emotional eating, and cognitive restraint. In contrast, there was no significant association between the amount of habitual sleep measured by wrist actigraphy and any of these eating behavior factors.

It has been argued that cognitively restrained eating is an adaptive behavior for individuals prone to gain weight in the setting of easy access to abundant and palatable food (35). However, young adults are highly susceptible to failure of restraint when it is associated with other potentially problematic eating behaviors, such as uncontrolled and emotional eating (36). Our findings that lower subjective sleep quality reflected by PSQI scores at the higher end of the typical range seen in good sleepers (26) are associated with increased hunger, uncontrolled and emotional eating, and more cognitive restraint suggest that small decrements in self-reported sleep quality can be a sensitive indicator for the presence of potentially problematic eating patterns in young urban adults at high risk for type 2 diabetes. Self-reported sleep quality did not reflect specific changes in the architecture, consolidation and efficiency of sleep in the laboratory (assessed by polysomnography) or the habitual amount and fragmentation of sleep at home (measured by wrist actigraphy), but was related in part to individual ratings of depressed mood. Consistent with our findings, population-based observations in the Penn State (19) and MONICA/KORA (18) cohort studies indicate that complaints of poor sleep quality and psychological distress could be important determinants of the association between self-reported short sleep and chronic metabolic morbidity. In the Penn State cohort for example, complaints of poor sleep and measures of psychological distress were the primary predictors of self-reported short sleep among obese participants: those with insomnia reported having the shortest sleep duration averaging 5.9 h per night, whereas the subjective sleep duration of obese individuals without sleep complaints (7.0 h) was similar to that of non-obese good sleepers (6.9 h) (19). Therefore, future investigations of the relationship between sleep and the control of human eating behavior, hunger, and body weight regulation should include formal screening for the presence of sleep pathology and assessment of the psychological well being of the study participants (19).

Consistent with previous population-based data (37), nearly 40% of our subjects habitually slept < 6 h/night. Prior laboratory experiments have found that young men exposed to sleep curtailment of 4 h/night and caloric restriction (1500 kcal/day for the average 75 kg study participant) at the time of sampling have lower circulating concentrations of the anorexigenic hormone, leptin, higher concentrations of the orexigenic hormone, ghrelin, and increased subjective hunger (8). Supported by observations from the Wisconsin Sleep Cohort showing a positive association between leptin and self-reported sleep time, and an inverse association between ghrelin and polysomnographic sleep time (4), these findings have given rise to the widespread notion that insufficient sleep triggers key hormonal signals of “famine in the midst of plenty” to cause excessive food intake and weight gain.

More recent experiments have exposed human volunteers to sleep restriction in the presence of adequate or excess amounts of self-selected calories. In these studies, short-term sleep restriction was accompanied by increased leptin concentrations in women (9, 13–15) and had no independent effect on leptin in men (10, 16). Other experiments combining two weeks of sleep restriction with over- or underfeeding found that sleep loss did not interfere with the expected physiological rise or fall in leptin concentrations, while sleep-loss-related increases in ghrelin levels and hunger were seen only in the presence of negative, but not positive, energy balance (11, 12). These experimental findings and some newer epidemiological data (17) suggest that prior reports of increased hunger, lower leptin and higher ghrelin concentrations related to short-term sleep restriction (8) did not reflect the presence of “famine in the midst of plenty”, but rather the ability of sleep loss to amplify the human behavioral and neuroendocrine response to caloric restriction (12). In agreement with these newer data on the relationship of hunger-regulating hormones with sleep loss (9–11, 13–17), there was no significant association between measured sleep duration and hunger-related eating in the present study. Similarly, the amount of habitual sleep was not a significant predictor of any other self-reported eating behaviors in free-living adults with familial risk for type 2 diabetes.

Physical activity can modulate eating behavior and the relationship of TFEQ scores with measures of adiposity (38). Some epidemiologic (28, 29, 31) and experimental studies (10) also indicate that short sleep is accompanied by lower levels of habitual physical activity. In the present study, higher amounts of objectively-measured physical activity were associated with a more uncontrolled and hunger-dependent eating behavior (Table 2). As reported elsewhere (24), there was also a positive association between habitual sleep duration and the amount of everyday physical activity in the participants of this study. However, inclusion of sleep duration and physical activity measured by wrist actigraphy and waist accelerometry as independent variables in the multiple regression analysis did not attenuate the significant association of self-reported sleep quality with important aspects of eating behavior (Table 2). Given the association between sleep time and free-living physical activity (10, 24, 28, 29, 31), and between free-living physical activity and uncontrolled and hunger-dependent eating behavior (Table 2), future studies of the link between sleep and eating behavior should take into consideration the large individual differences in the amount of everyday physical activity.

Our study has several strengths and limitations. We collected a set of exploratory data using a carefully screened sample of healthy individuals at high risk for type 2 diabetes, while avoiding the potentially confounding effects of obesity and its co-morbid conditions on various aspects of eating behavior. The use of laboratory polysomnography and continuous ambulatory monitoring of habitual sleep and free-living physical activity also allowed us to exclude the presence of sleep pathology and avoid assessments based on unreliable self-reports of these behaviors. Finally, it was important to study a population with high risk for type-2 diabetes which may inform future behavioral research on sleep and metabolic risk reduction. Despite its strengths, this was an exploratory study which included a relatively small number of subjects who were not randomly selected and the results may not be entirely representative of the relationship between sleep and eating behavior in this population. Furthermore, all assessments of eating behavior were based on self-report and we do not know what are the implications of the link between subjective sleep quality and key eating behavior factors for the long-term regulation of energy intake and body weight in this high-risk population.

In conclusion, our results suggest that small decrements in self-reported sleep quality can be a sensitive marker for the presence of eating patterns characterized by increased hunger, uncontrolled and emotional eating, and cognitive restraint – a potentially problematic combination (36) - in young urban adults with increased risk of developing type 2 diabetes. Lower self-reported sleep quality did not reflect changes in sleep structure and consolidation assessed by wrist actigraphy and polysomnography, but was correlated in part with individual ratings of depressed mood. Since emotional distress and problematic eating patterns can be ameliorated, complaints of reduced sleep quality in such susceptible individuals may warrant special attention when considering specific lifestyle modification strategies for metabolic risk reduction. Additional studies are needed to explore the association of reduced sleep quality, emotional well being, and problematic eating behaviors in individuals with increased susceptibility to obesity and type 2 diabetes.