No effects of short-term sleep restriction, in a controlled feeding setting, on lipid profiles in normal weight adults

Summary

Short sleep has been associated with cardiovascular risk. The aim of this study was to determine the impact of short-term sleep restriction on lipid profiles and resting blood pressure factors in young, normal weight individuals (14 men, 13 women). Participants were randomized to 5 nights of either habitual (9 h) or short (4 h) sleep in a crossover design separated by a 3 wk washout period. There was no sleep x day interaction on lipid profile and blood pressure. Short-term sleep restriction does not alter lipid profiles and resting blood pressure in healthy, normal weight individuals. The association between short sleep and increased cardiovascular risk reported in the epidemiological literature may be the result of long-term sleep restriction and poor lifestyle choices.

Introduction

Restricting sleep duration increases food intake (St-Onge et al., 2011), decreases physical activity (Schmid et al., 2009) and up-regulates the stress response system (Omisade et al., 2010). Short sleep duration is also related to risk of adverse lipid profiles (Bjorvatn et al., 2007) but data are equivocal. In one clinical study, total (TC) and low-density lipoprotein cholesterol (LDL-C) levels in post-menopausal women increased following 3 nights of 4 h/night in bed compared to 8 h (Kerkhofs et al., 2007).

The main study tested the hypothesis that bedtimes to 4 h/night, relative to 9 h/night, would lead to a metabolic profile indicative of positive energy balance with propensity for weight gain. The secondary hypothesis was that restricting sleep would result in an unfavorable lipid profile.

Materials and Methods

Men and women, age 30 – 45 y, BMI 22 – 26 kg/m and regular sleep duration 7 – 9 h/night, were recruited via approved media. The study was approved by the Institutional Review Boards of St. Luke’s/Roosevelt Hospital and Columbia University, and informed consent was obtained prior to enrollment (Clinical trials #NCT00935402).

This study was a randomized, controlled, crossover study with 2 phases of 6 d (St-Onge et al., 2011) with either habitual (time in bed 2200-0700 h) or short sleep (0100–0500 h). Sleep phases were separated by 3 wk to ensure that participants returned to their pre-study sleep schedules for phase 2 and that women were tested in the same phase of their menstrual cycle. During the additional awake time associated with the short sleep phase, participants were ambulatory and exposed to normal indoor lighting (~500 lux).

For the first 4 d of each phase, participants consumed a controlled diet (30% of energy from fat, 15% protein, and 55% carbohydrates). Meals (30% of energy requirements)and snack (10% of energy)were served at fixed times: 0800 (breakfast), 1200 (lunch), 1600 (snack), and 1900 (dinner). Food intake was ad libitum d 5–6. Fasting blood samples, blood pressure and resting heart rate (RHR) were taken daily. On d 4, blood samples were taken at 0800, 0815, 0830, 0900, 0930, 1000 h and every 2 h thereafter until 0600h on d 5.

Fasting TC, high-density lipoprotein cholesterol (HDL-C) and triglycerides (TG) were assessed by spectrophotometry (Ektachem DT 11 System, Johnson and Johnson Clinical Diagnostics, Rochester, NY) with appropriate reagents and standards (Cliniqa Raichem, San Marco, CA). LDL-Clevels were calculated (Friedewald et al., 1972).

Statistical Analysis

Comparisons were made between samples collected at baseline and d 5 using linear mixed model analysis. With more than 100 data points, normal approximation of t-test and F-test statistics were used to perform the analysis, in accordance with statistical large sample theory. Log-transformation did not improve the normal approximation of any variable (P-value > 0.05). We used the response measures as the dependent variable in each linear mixed model analysis. Sleep (short vs habitual), day (1 vs 5), body weight and sex were used as fixed effects and a sleep-by-day interaction was tested. We took into account the repeated measure-structure of the data by treating participants as a random effect. In the case of repeated measurements of the same variable in a single day (TG 24 h analysis), we also used time (morning vs afternoon vs night) as a fixed effect in the combined analysis. Three separate linear mixed model analyses for morning (0800 – 1200 h), afternoon (1230 – 1900 h) and night (1930 – 0600 h) were also conducted for 24-h TG data. Analyses were repeated for men and women separately.

Data are expressed as means ± SD. A P-value of <0.05 was considered statistically significant.

Results

Fourteen men and 13 women completed the study (Table 1). One man and one woman were withdrawn due to uncovered exclusionary factors after enrollment (sleep disorder and anti-depressant medication use). Another woman voluntarily withdrew from the study after the first phase. During the short sleep phase men slept 225.5 ± 2.3 min/night while women slept 226.9 ± 1.0 min/night (St-Onge et al., 2011). Body weight decreased during both phases (habitual 1.09 ± 0.18 kg; short 0.77 ± 0.18 kg, both P<0.0001).

Table 1.

Characteristics of study participants.

Characteristics	Men	Women
Age, y	36.6 ± 5.6	33.9 ± 4.3
Weight, kg	77.1 ± 9.0	62.6 ± 5.0
Height, cm	178.6 ± 6.4	164.8 ± 6.5
BMI, kg/m	24.3 ± 1.5	23.0 ± 1.1

Data are means ± SD. BMI, body mass index. Men n=14, women n=13.

There was no sleep-by-day interaction or main effect of sleep on d 5 for any of the fasting lipids (Table 2). There was no effect of sleep on 24-h (P=0.65), morning (P=0.27), afternoon (P=0.94) and overnight (P=0.97) TG (Figure 1). There was a significant effect of sex (P<0.001) on 24-h TG but responses did not differ by sleep when analyses were done separately in men (P=0.73) and women (P=0.62). Morning, afternoon and overnight TG profile did not differ by sleep in men or women (Table 3).

Table 2.

Effect of sleep on fasting lipid profiles, blood pressure and resting heart rate following 4 nights of habitual (9 h/night) and short (4 h/night) sleep.

Risk Factor	Habitual Sleep		Short Sleep		P-value
	Day 1	Day 5	Day 1	Day 5
TC, mg/dLa	181.2 ± 41.0	171.0 ± 35.9	173.7 ± 37.9	161.4 ± 34.8	0.16
LDL, mg/dL	119.2 ± 39.0	108.3 ± 31.7	113.3 ± 40.5	100.8 ± 29.6	0.29
HDL, mg/dL	51.6 ± 16.1	48.3 ± 14.0	50.3 ± 14.5	47.5 ± 13.5	0.69
TG, mg/dL	66.3 ± 35.3	71.9 ± 25.5	73.7 ± 39.5	65.8 ± 26.0	0.90
SBP, mmHg	113.9 ± 7.8	107.2 ± 7.3	109.6 ± 7.8	107.7 ± 7.8	0.07
DBP, mmHg	73.3 ± 7.3	68.2 ± 6.2	71.3 ± 7.3	69.1 ± 7.8	0.50
RHR, bpm	62.0 ± 10.4	60.7 ± 12.5	66.8 ± 12.5	61.9 ± 8.8	0.10

Data are raw means ± SD, n=27. The P-value reflects a main effect of sleep in a linear mixed model analysis adjusting for the effect of Day (Day 1 vs. Day 5), and thus adjusting for baseline differences as well.

^aAbbreviations used: DBP, diastolic blood pressure; HDL, high-density lipoprotein cholesterol; LDL, low-density lipoprotein cholesterol; RHR, resting heart rate; SBP, systolic blood pressure; TC, total cholesterol; TG, triglycerides.

Figure 1.

Mean 24 h triglyceride (TG, mg/dL) profile following 3 nights of habitual (9h, black line) or short (4 h, broken line) sleep. The sleep episode is depicted by the back (9h) or broken (4h) line by the x-axis. Data are means ± SD, n = 27. Meals, as indicated, were served after the blood draw at 0800, 1200, 1600 (snack), and 1900 h. Habitual sleep time period: 2200 – 0700; short sleep period: 0100 – 0500.

Table 3.

Effect of short sleep on differences between adjusted means for 24 h, morning, afternoon and night-time triglycerides in men and women following 3 nights of short sleep (4 h/night) compared to habitual sleep (9 h/night).

	Men		Women
	Difference	P-value	Difference	P-value
24 h (0800-0600)	−1.49 ± 15.7	0.73	−1.29 ± 9.4	0.62
Morning (0800–1200)	−2.77 ± 10.5	0.55	−3.92 ± 14.4	0.33
Afternoon (1230–1900)	4.16 ± 46.0	0.74	−3.55 ± 22.4	0.57
Night (1930-0600)	−2.85 ± 30.3	0.20	3.39 ± 15.1	0.42

Data are adjusted mean difference ± SD for the difference between short and habitual sleep. The data were analyzed by linear mixed model analysis and adjusted for body weight. Men n=14; women n=13.

Effects of sleep on systolic and diastolic blood pressure and RHR were not significant (P=0.73, 0.47, and 0.64, respectively) on d 5.

Discussion

Short-term sleep restriction had no effect on lipid profiles in young men and women under conditions of mild negative energy balance.

Associations between unfavorable lipid profiles and reduced sleep have been reported in observational studies (Bjorvatn et al., 2007, Kaneita et al., 2008). In a clinical study, TC and LDL-C were elevated following 3 nights of 4 h sleep/night (Kerkhofs et al., 2007). Differences in study populations may account for the discrepant results: we enrolled healthy young adults whereas Kerkhofs et al. studied older, post-menopausal women on hormone replacement therapy (Kerkhofs et al., 2007). Effects of acute sleep restriction may be more pronounced in older individuals at metabolic risk.

Our study is the first to examine the effects of short-term sleep restriction on 24-h TG. Alterations in sleeping patterns modulate the TG profile and may explain the increase in overnight TG commonly seen in the shift worker population (Knutsson et al., 1986). We failed to confirm our hypothesis that the reduction in sleep duration would raise the 24-h TG profile, particularly in the overnight period.

Sleep restriction has been shown to be associated with increased blood pressure and hypertension (Robillard et al., 2011). In the present study, there was no interaction between sleep duration and day.

Dietary choices, rather than a direct effect of sleep, may present an alternative hypothesis to explain the relationship between short sleep and increased CVD risk. Increased energy intakes, particularly from fat and snack foods, have been noted during periods of restricted sleep in this same cohort and others (St-Onge et al., 2011, Nedeltcheva et al., 2009). It is possible that long-term sleep restriction resulting in poor dietary and lifestyle choices leads to increased cardiovascular risk. These changes would not be detected within the context of a controlled diet and mild energy restriction. An important factor that needs to be addressed when comparing the data from the current study to previous studies is energy balance and diet composition. Controlling for food intake does not guarantee energy balance. A poor diet and positive energy balance may contribute to the negative health effects of sleep restriction.

Another important factor to consider when interpreting the results of the current study is that of a potential circadian phase shift. Alterations in circadian rhythm are associated with metabolic dysfunction (Van Cauter et al., 2008, Lund et al., 2001, Buxton et al., 2012, Scheer et al., 2009). In the current study, participants were exposed to additional light stimuli during the short sleep phase. Since markers of circadian phase were not measured, it is difficult to determine the actual effect of the additional light exposure. However, the additional light exposure was distributed evenly between the night and morning and the nighttime phase delay may have been, in part, cancelled by the morning phase advancement. Nevertheless, measures of circadian markers and wakefulness EEG should be considered in future studies. Moreover, vigilance was not assessed throughout the study. Although participants were under constant supervision to ensure wakefulness, the presence of micro-sleep episodes cannot be ruled out.

In this study, participants consumed identical diets for both phases, removing a confounding effect of dietary intakes and uneven energy balance on the parameters measured. The inpatient setting provided a constant environment and conditions during both study phases. Finally, this study provides valuable information on the effects of sleep restriction in women, an understudied population.

Power analyses were performed based on paired t-test and linear mixed models, based on the TG differences in Kerkhofs et al. (2007) (approximate conservative effect size for t-test 0.53) as well as Cohen’s convention of effect sizes (t-test: small=0.2, medium=0.5, large=0.8; linear model: small=0.02, medium=0.15, large=0.35). Using our 100 data-points (27 subjects measured at least four times for each statistical analysis), we had at least 80% power to detect approximate t-test effect sizes derived from Kerkhofs et al. (2007), as well as Cohen’s medium and large effect size for t-test and linear mixed model. However, we had 51% power to detect Cohen’s small effect size of 0.2 for t-test and 15% power to detect Cohen’s small effect size of 0.02 for linear model analysis. However, the small effect sizes are not as relevant as Kerkhofs et al. (2007) reported a medium effect size.

Finally, our results do not support the hypothesis that short-term, reduced sleep duration negatively affects the lipid profile, specifically in the context of negative energy balance and a healthy diet. The sustained exposure to long-term sleep restriction and subsequent poor lifestyle choices may explain the relationship between short sleep and cardiovascular risk in the epidemiological literature. These considerations should be examined in future studies.