Introduction

In recent decades, there has been increased interest in understanding the potential pathway, whereby poor sleep quality may affect obesity risk, particularly in the context of a worldwide obesity epidemic. However, the vast majority of the literature on this topic is focused on populations of children and younger adults. This is particularly concerning given that Census estimates indicate that by 2030, one in five persons in the U.S. will be over the age of 65 [1].

As with other psycho-biological processes, sleep patterns substantially change across the lifespan with important differences between objective and subjective sleep. A meta-analysis of 65 studies investigating changes in sleep architecture across the lifespan and using polysomnographic recordings showed that sleep latency (i.e., the time needed to fall asleep) as well as light sleep increase with age, while slow-wave sleep and REM sleep decrease [2]. Importantly, sleep becomes more fragmented with age, as the number of awakenings as well as the time spent awake at night increase [2, 3]. These objective impairments, though, seem only partially reflected in subjective reports. In fact, there is evidence from population based studies that sleep complaints are less present in older adults than in younger individuals [4]. In summary, several normal sleep changes occur with aging, including robust evidence that sleep becomes lighter and fragmented with age. Nonetheless, studies on subjective sleep quality reported mixed results [4, 5]. Strong predictors of poor sleep in the elderly are: female gender, mood disorders, and physical illness [6, 7].

Alongside normal changes in sleep patterns, older adults are also vulnerable to several sleep disorders. Though not a focus of this review, we briefly mention here that older adults may also be more prone to insomnia [5], sleep apneas [8], and sleep-related movement disorders [9, 10] than younger adults. Such sleep disorders can exacerbate sleep fragmentation and other normal age-related sleep changes.

Because of the relative scarcity of research on the health effects of sleep quality on obesity risk in late life, the aim of the present narrative review is to examine the literature studying older adults to describe the association between poor sleep quality and obesity. We hypothesize, and the literature supports, the causal direction of poor sleep increasing risk for obesity. However, it should be noted that there are some studies finding evidence for the reverse pathway, where obesity was causally linked to later development of insomnia, difficulties maintaining sleep, and excessive daytime sleepiness [11]. Finally, in light of the mounting evidence summarized in this review that poor sleep may be a serious risk factor for obesity in older adults, we discuss the urgency to identify targets for intervention that may interrupt this unhealthy pattern through positive changes in modifiable lifestyle factors that show promise in improving sleep quality and quantity.

Method

An online search on PubMed and PsycInfo databases was conducted by the authors, and subsequently, a hand search of the reference lists of all papers meeting the inclusion criteria was also performed.

The search keywords included the following: “bmi” OR “body mass index” OR “obesity” OR “overweight” AND “elderly” OR “aged” OR “older” OR “elder” OR “geriatric” AND “sleep”.

To be included in this review, a quantitative study had to meet all the following inclusion criteria:

  1. a.

    should have been published in peer-reviewed journals during or since 1990;

  2. b.

    should have focused on older population (60+);

  3. c.

    should have reported at least a relation between sleep and BMI;

  4. d.

    should have used objective or self-reported measures;

  5. e.

    should have been published in English.

Papers were excluded if:

  1. a.

    they target full range of adulthood (18+) except those with stratified/separate results for older adults;

  2. b.

    they focus on special medical populations;

  3. c.

    they discuss specific sleep disorder.

Since this is a narrative review, no other guidelines were followed during the search and the selection of the studies. Given that the majority of studies published to date utilize subjective measures of both sleep quality and self-reported height and weight, these are included in the present review. However, we separately report on conclusions drawn from objectively measured sleep quality and anthropometrically measured obesity where available, and emphasize the greater internal validity of such findings and the extent to which they corroborate or refute results from studies relying on subjective measures. While the vast majority of studies are cross-sectional, also included in this review are the more rare longitudinal studies that provide stronger causal evidence.

The most common measure of sleep quality is sleep duration, but we also consider studies providing additional metrics including sleep latency, perceived quality of sleep, daytime sleepiness/napping, percentage of time in rapid eye movement (REM) sleep, and nocturnal awakenings. A discussion then follows which considers psychological moderators affecting the manner in which depression, anxiety, and perceived stress may affect the likelihood that poor sleep would lead to obesity. We then we also look at the most widely accepted biological mediators and moderators, with a description of the physiological mechanisms that make these biological factors so salient.

Results

A summary of the main characteristics of the studies included in this narrative review is provided in Table 1. Below, we first review findings from cross-sectional studies, acknowledging that they do not provide evidence of causation. However, given the paucity of longitudinal studies, it is important to examine the extent to which cross-sectional findings are consistent with a causal pathway.

Table 1 Methodological features of the studies included in the present literature review
Full size table

Cross-sectional studies of subjectively measured sleep quality

Goerke et al. [12] in a pilot study investigating the role of habitual sleep duration in physical exercise-induced weight loss reported an inverse association between body mass index (BMI) and mean sleep duration in a sample of 22 community-dwelling healthy elderly aged 61–76 years. In the cross-sectional analyses at baseline, overweight subjects (anthropometrically measured BMI > 25 kg/m2) reported a 50-min shorter sleep duration than normal weight subjects. Furthermore, subjects who slept less than 7.5 h per night had a 3.13 kg/m2 higher BMI compared to subjects who slept more than 7.5 h per night.

A cross-sectional study by Gildner et al. [13] pooled nationally representative samples of older adults (age > 50 years) in six middle-income countries. In models stratified by gender and adjusting for lifestyle factors, self-reported longer sleep duration was significantly associated with lower anthropometrically measured BMI (β = − 0.058; p < 0.001) and smaller waist circumference (WC; β = − 0.063; p < 0.001) in men; and similarly, lower BMI (β = − 0.076; p = 0.039) and smaller WC (β = − 0.086; p = 0.030) in women. Ford et al. [14] found that self-reported sleep duration showed inverse linear associations with body mass index and waist circumference (measured anthropometrically) in a large nationally representative sample from the National Health and Nutrition Examination Survey 2005–2010 (NHANES). The entire sample showed this association, including the subgroup of participants aged ≥ 60 years. These various studies’ results suggest that short sleep duration is potentially harmful for physical well-being in the elderly.

There is also evidence of a curvilinear relationship between sleep and BMI, and specifically of the maladaptive role of long sleep duration. For example, Littman et al. [15] studied 173 post-menopausal sedentary overweight women aged 50–75 years and found that higher BMI was weakly associated with longer sleep duration (p trend = 0.06) and more frequent daytime napping (p trend = 0.08). Tu et al. [16] studied 68,832 women aged 45–80 years who participated in the Shanghai Women’s Health Study. The odds of experiencing ≤ 4, 5, 6, 8, 9, or ≥ 10 h of sleep per night (compared to 7 h, the reference category) were modeled in polynomial logistic regressions. Higher BMI was associated with greater odds of longer sleep (8 and ≥ 10 h; p trend = 0.001 for both) and lower odds of shorter sleep (≤ 4 and 5 h; p trend = 0.001 for both). Similar results were reported for higher waist–hip ratio (WHR) and WC. Thus, obesity is also linked to a longer sleep duration.

A possible explanation for disagreement in the above findings—studies showing obesity linked to both shorter and longer sleep durations—is that both could be true. A number of studies have reported a U-shaped distribution, where especially short or especially long sleep duration was a risk factor for obesity. Liu and colleagues [17] report on data from 54,269 adults age 45 years and older who completed the 2010 Behavioral Risk Factor Surveillance System survey in 14 states in the US. After controlling for demographic covariates, OR (95% CI) for obesity (self-reported BMI ≥ 30 kg/m2) were 1.32 (1.21–1.43) for six or fewer hours, 1.00 (referent) for 7–9 h, and 1.60 (1.34–1.93) for ten or more hours of sleep per night. These results suggest that short and particularly long sleep durations are unhealthy.

Another U-shape pattern was reported by Tuomilehto et al. [18] who examined 2770 older adults (aged 45–74) from the FIN-D2D (Finnish) study. Prevalence of overweight (anthropometrically measured BMI 25–30 kg/m2), obesity (BMI > 30 kg/m2), or central obesity (WC exceeding World Health Organization or WHO criteria) varied by sleep duration. Subjects with 6 or fewer hours or 8 or more hours were less likely to be overweight (compared to 7 h; p = 0.035), and more likely to be obese (p = 0.003) and to have central obesity (p < 0.001).

Magee et al. [19] analyzed data from 45,325 Australian adults (aged 55–95) in regression models. Self-reported short sleep was associated with an increased risk of obesity in the entire sample, as was long sleep duration, evidencing another U-shaped association between sleep duration and obesity. When the analyses were stratified by age, short and long sleep durations were associated with obesity in 55–64 years but not in those aged 65 years and above. Similarly, Ohayon and Vecchierini [20] studied 1026 randomly selected French older adults aged 60 years and above (76% of whom were aged 65 years and above) to examine BMI characterized as extreme value (i.e., in the fifth lower and upper percentiles of the distribution). Habitual bedtime exhibited a U-shaped distribution with obesity (defined as BMI > 27) with both early (≤ 9 p.m.) and late (≥ 1 a.m.) bedtime associated with higher obesity risk. Obesity was also associated with night-time short sleep duration (≤ 4 h 30; OR = 3.6, 95% CI 1.0–13.1) and late wake-up time (≥ 9 a.m.; OR = 2.2, 95% CI 1.3–4.0). Moreover, a daytime sleep duration of 1 h or more (95th percentile; OR = 1.8, 95% CI 1.1–3.0) was also associated with obesity. Finally, Gildner et al. [13] analyzed data from the first wave of the WHO Study on Global Aging and Adult Health. Longer sleep duration was associated with lower BMI in both men (β = − 0.058; p < 0.001) and women (β = − 0.076; p = 0.039). However, higher ratings on a 2-day average self-reported sleep quality report were associated with higher BMI in men (β = 0.042; p = 0.001) but not in women (β = − 0.008, p = 0.804).

While the above studies primarily examined sleep duration, additional evidence suggests that other aspects to sleep quality are also linked to obesity. Vaatainen et al. [21] studied self-reported nocturnal awakening in a sample of 823 community-dwelling older adults in Finland (aged 55–75). “Frequent” (often or very often) vs. “infrequent” (never, rarely, and sometimes) nocturnal awakening was associated with higher self-reported BMI (M = 27 ± 0.30 kg/m2 vs. M = 26.6 ± 0.17 kg/m2; p = 0.009).

In another study of 702 older adults from the community-based Einstein Aging Study of aging (aged M = 80 ± 5.5 years), sleep difficulties were coded as “moderate/severe” (SSD, “most” or “all of the time”), “mild” (MSD, “a good bit” or “some of the time”), or “never” (NSD) from self-reporting over the past 4 weeks as having had trouble falling asleep and/or problems with awakening during sleep. Obesity (BMI ≥ 30 kg/m2) rates per self-report differed significantly between the three sleep difficulty groups (SSD = 32.5%, MSD = 26.0%, and NSD = 21.8%, p < 0.03) [22]. Jaussent et al. [23] found in a sample of 5886 community-dwelling older adults (aged 65+) that self-reported high BMI (> 30 vs. < 25) predicted nearly triple the risk of self-reported difficulty initiating sleep but only in men (OR = 2.7, p = 0.036).

Recently, Moreno-Vecino et al. [24] studied 463 community-dwelling older Spanish women (aged 66–91), coding sleep disorder according to DSM-IV-TR criteria, as difficulty initiating/maintaining sleep or non-restorative sleep at least three nights per week for at least 1 month. The group with sleep disorders showed significantly greater values of self-report-derived BMI (30.0 ± 4.4 vs. 28.9 ± 4.2 kg/m2, p < 0.05, with and without sleep disorders, respectively), and WC (94.1 ± 11.5 vs. 91.3 ± 10.3 cm, p < 0.01, with and without sleep disorders, respectively) than those without sleep complaints.

Frequent daytime sleepiness predicted higher self-reported BMI, compared to never/rare daytime sleepiness (M = 28.70 ± 6.9 vs. M = 26.7 ± 5.3, respectively, p < 0.001) in a cross-sectional study of 1506 persons (aged 55–84) from a national US survey [25]. Sforza et al. [26] examined 825 healthy elderly (aged 65 years and older), measuring excessive daytime sleepiness (EDS) defined on the basis of the Epworth Sleepiness Scale score ≥ 10. BMI differed significantly in those coded as “sleepy” vs. “not sleepy” per EDS, with averages of 26.3 ± 3.6 and 25.3 ± 3.8, respectively (p = 0.007).

In a sample of 5201 older adults, aged 65 years and older who were participants in the Cardiovascular Health Study, Enright et al. [27] found that the following self-reported sleep disturbances were all significant predictors of higher self-reported BMI: loud snoring (p < 0.001 for males and females), observed apneas (p = 0.002 for males, p = 0.005 for females), and daytime sleepiness (p < 0.001 for males, but n.s. for females).

It is important to note also that a number of studies failed to find a significant association between sleep duration or quality and obesity. For example, in a sample of 5607 participants of the NHANES, Grandner et al. [28] found that self-reported sleep duration was a significant predictor of anthropometrically measured BMI but only in younger age groups (age < 50 years); however, it was non-significant in both 50–64 years and in those aged 65+, before and after covariate adjustment. Authors speculate that survival bias may at least partially explain attenuated results among older adults or that effects of sleep on body weight may have already manifest at younger ages. Similarly, a study by Gangwish et al. [29] examined data from the 1982–1984 NHANES I follow-up survey. Authors evaluated the cross-sectional association between self-reported sleep duration and obesity status (Lean BMI ≤ 25, Overweight BMI > 25 and < 30, Obese BMI ≥ 30). Although a significant association was found in younger subjects aged 32–49 (χ 2 = 55.34, p < 0.0001), no such association was found for subjects aged 50–67 years (χ 2 = 19.41, p = 0.0792) or 68–86 years (χ 2 = 11.96, p = 0.4486). In a sample of 624 older adults (aged 60–95), McHugh et al. [30] reported no association between BMI and perceived sleep quality, measured by the Pittsburg Sleep Quality Index-PSQI (good: PSQI < 5 vs. poor: PSQI ≥ 5) (p = 0.223).

Chien et al. [31] examined 488 community-dwelling older adults (aged 65+) and found that sleep duration (trichotomized into < 6, 6–8, and ≥ 8 h/night) was not significantly associated with BMI (p = 0.23 and p = 0.32 in males and females, respectively), obesity (BMI > 27, p = 0.58 and p = 0.15 in males and females, respectively) or percent body fat (p = 0.15 and p = 0.95 in males and females, respectively). Similarly, Gottlieb et al. [32] studied a sample of 1486 subjects (age M = 70.2 ± 8.5), and found no difference in BMI among different sleep duration categories (≤ 5, 6, 7–8, and ≥ 9 h of habitual sleep).

Chaput et al. [33] assessed anthropometrics and body composition, resting energy expenditure, daily energy expenditure, daily energy intake, plasma lipid–lipoprotein profile, and self-reported sleep duration in 90 women (age M = 60.9 ± 6.6). In all the variables, no significant differences were found between two groups based on sleep duration. Women reporting ≥ 7 h of sleep had a similar OR for obese/overweight compared to those reporting short sleep duration (< 7 h). Another study [34] explored the effect of sleep duration on 5-year abdominal fat accumulation in a two minority groups of African (N = 332) and Hispanic (N = 775) Americans (aged 18–81), finding no significant relationship between sleep duration and fat deposit changes in participants older than 40 years.

Longitudinal studies of subjectively measured sleep quality

Longitudinal studies provide stronger evidence for a causal linkage between poor sleep and obesity, though the number of such studies is small. Xiao et al. [35] in a large investigation in the NIH-AARP Diet and Health Study Cohort with 83,377 participants (aged 51–72), prospectively explored the association between self-reported sleep duration and weight change over an average of 7.5 years of follow-up (1995–2004). Short sleep duration (< 5 and 5–6 h) was associated with 0.66 kg (95% CI 0.19, 1.13) and 0.43 kg (95% CI 0.19, 1.13) more weight gain compared with normal sleep duration (7–8 h) with p for trend = 0.02 and p for trend < 0.001, for men and women, respectively. Short sleepers (< 5 h) had an approximately 40% higher risk of developing obesity than normal sleepers (7–8 h) (for men, odds ratio = 1.45, 95% CI 1.06, 1.99; for women, odds ratio = 1.37, 95% CI 1.04, 1.79).

Nagai et al. [36] followed up on 13,629 Japanese participants (aged 40–79) prospectively for up to 11 years, and used logistic regression to predict weight gain of at least 5 kg, or being obese (self-reported BMI ≥ 25) at follow-up, as a function of baseline self-reported sleep duration. Overall, sleep duration did not significantly predict either outcome, but among participants who were obese at baseline, those sleeping for 9+ h per night were at increased risk for ≥ 5 kg weight gain, with OR = 1.36 (1.09–1.70).

A study conducted by Lopez-Garcia et al. [37] examined the cross-sectional and longitudinal association between habitual sleep duration and obesity, abdominal obesity, and weight change in 3576 older adults (aged 60+) in Spain. Participants underwent a home-based personal interview and physical examination in 2001 and a telephone interview in 2003. Individuals reporting ≤ 5 h of sleep had a greater frequency of obesity (odds ratio, OR 1.33; 95% CI 1.00, 1.77) and severe obesity (OR 2.08; 95% CI 1.31, 3.32) compared to individuals reporting 7 h. Surprisingly, a sleep duration of 8 h was associated with obesity (OR 1.39; 95% CI 1.11, 1.75) and severe obesity (OR 1.82; 95% CI 1.21, 2.73) as well as 9 h of sleep that was associated with severe obesity (OR 1.57; 95% CI 1.00, 2.47). The same pattern of results was found in the association with weight change over time among women but not men. A weight gain of 5 kg was more frequent among women reporting ≤ 5 h (OR 3.41; 95% CI 1.34, 8.69), 8 h (OR 3.03; 95% CI 1.29, 7.12), and 9 h (OR 3.77; 95% CI 1.55, 9.17) of sleep. No results were found between sleep duration and abdominal obesity.

Objectively measured sleep quality

Most of the studies reported in this review provide results from self-reported measurements of sleep quantity or quality, which is subject to recall bias. For this reason, we also present studies employing objective measures of sleep. To the extent that studies using objective measures yield results that mirror those from subjective sleep quality, validity of the latter is strengthened. A sample of 221 overweight or obese women (aged 40–69) was studied by Shade et al. [38] over a 6-month interval. BMI and WC were anthropometrically measured, subjective sleep quality per interview, and via actigraphy quantified total sleep time (TST), number of awakenings, wake after sleep onset (WASO), and percent sleep. There were weak positive associations between the change in self-reported sleep disturbance and change in weight (r = 0.202, p < 0.05), BMI (r = 0.211, p < 0.05), WC (r = 0.169, p < 0.05), and diastolic blood pressure (r = 0.137, p < 0.05). In contrast, change in total sleep time was not associated with change in BMI or weight, but weak positive associations were found between the number of awakenings and changes in weight (r = 0.167, p < 0.05), BMI (r = 0.171, p < 0.05), WC (r = 0.173, p < 0.05), and both systolic (r = 0.175, p < 0.05) and diastolic (r = 0.208, p < 0.05) blood pressures.

Recently, Cross et al. [39] evaluated napping habits in older adults at risk of dementia (age M = 65.5 ± 8.4) using subjective (sleep diaries) and objectively (actigraphy) measures of sleep. Participants also underwent comprehensive medical, psychiatric, and neuropsychological assessment. Results showed that higher rates of napping were significantly associated with greater BMI (r = 0.209, p < 0.05) and medical burden (r = 0.193, p < 0.05) and higher rates of mild cognitive impairment (i.e., impaired verbal fluency, r = 0.205, p < 0.05). Other findings of Anderson et al. [40] demonstrated that subjects with abnormal sleep–wake cycles (assessed over 5–7 days with a novel wrist triaxial accelerometer) had higher BMI than those with normal sleep–wake cycle (p = 0.002) in a cohort study of 421 very old men and women aged 87–89.

Blackwell et al. [41] reported on findings from a sample of 2909 community-dwelling men (aged 67+). Using in-home polysomnography, the percentage of time in REM sleep was recorded and grouped into quartiles; however, there were no differences found in BMI between the four groups (p = 0.68).

Another study using objective measurements of sleep duration and fragmentation was conducted by van den Berg et al. [42] using a six night actigraphic assessment and anthropometrics measurement to calculate BMI. A marked U-shaped association of actigraphic measures of sleep duration with BMI and obesity was reported (b of quadratic term = 0.30, 95% CI 0.08, 0.52). Compared to participants sleeping 7 to < 8 h, those sleeping < 5, 5 to < 6 or ≥ 8 were more likely to be obese (OR 2.76, 95% CI 1.38, 5.49; OR 1.97, 95% CI 1.26, 3.08; OR 2.93 95% CI 1.39, 6.16, respectively). This association was not found after adjustment for sleep fragmentation, whereas the higher risk for long sleepers remained unchanged. However, a quadratic relationship between sleep duration and BMI still existed after adjustment for sleep fragmentation. The latter also increased the likelihood of a higher BMI and obesity.

A study by Patel and colleagues [43] evaluated the association between sleep duration and weight and body composition (objectively assessed) in two cohorts of older adults. A sample of 3055 men (aged 67–96) participating in the Osteoporotic Fractures in Men Study (MrOS) and 3052 women (aged 70–99) participating in the Study of Osteoporotic Fractures (SOF) wore a wrist actigraphy for a mean of respectively 5.2 ± 0.9 nights and 4.1 ± 0.8 nights. Sleep duration of < 5 h was associated with BMI that was on average 2.5 kg/m2 (95% CI 2.0, 2.9) greater in men and 1.8 kg/m2 (95% CI 1.1, 2.4) greater in women, compared to a sleep duration of 7–8 h. Short sleep was associated with central body fat distribution and increased percent body fat even controlling for sleep apnea, insomnia and daytime sleepiness. Moreover, the odds of obesity were 3.7-fold greater (95% CI 2.7, 5.0) in men and 2.3-fold greater in women (95% CI 1.6, 3.1) sleeping less than 5 h.

Patel et al. [44] also examined a sample of 3053 men (age M = 76.4) and 2985 women (age M = 83.5) using wrist actigraphy. Results showed positive associations between variability in nightly sleep duration and daytime napping and obesity. Particularly, with increasing variability in sleep duration and daytime napping, they found an increase in the mean BMI in both men and women with highly significant tests for trend (p ≤ 0.001), even after adjustment for mean nocturnal sleep duration. Another study by Rao and colleagues [45] utilized a sample of 2745 older men from the Osteoporotic Fractures in Men Study (MrOS) who underwent polysomnography. Percentage of time in slow-wave sleep (SWS) was evaluated, a state of sleep important for “sleep-dependent memory processing” [46]. Using multiple linear regression analysis, they found that percentage of time in SWS was inversely associated with BMI (p trend = 0.0095) and other measures of body composition as WC (p trend = 0.001) and weight (p trend = 0.089), even after adjusting for potential confounding variables.

Kim et al. [47] in a cross-sectional study including 191 Japanese community-dwelling women (aged 80–92) examined whether night-to-night sleep variations were associated with body composition. Authors did not find an association between objectively measured sleep duration variability and BMI in very elderly women. However, sleep duration variability was associated with body composition indices of obesity, positively with the percentage body fat mass (r = 0.22, p < 0.01) and negatively with the percentage of lean mass (r = − 0.21, p < 0.01).

In summary, studies using objective measurement of sleep quality yield findings that are generally consistent with those from studies using self-report measures, lending validation to the self-report studies. However, it should be noted that all of the objectively measured sleep quality studies were cross-sectional and, therefore, are inconclusive with respect to causal pathway. We next turn attention to important psychological and physiological mediators and moderators of the sleep/obesity linkage to more fully explicate this complex association.

Psychological mediators and moderators

Psychological conditions such as depressive symptom, anxiety and stress can seriously impact sleep quality and sleep duration. At the same time, sleep deprivation and fragmentation can increase the risk of psychological distress. It can also directly impact weight gain due to its effect on eating behaviors, and if not resolved, the eating behavior pattern may worsen the problem of sleep deprivation and produce unwanted changes in body weight among other factors. There is a complex bidirectional relationship between several different sleep disorders and obesity [48]. As such, although not commonly cited as a primary determinant of obesity, sleep deprivation, restriction, or sleep fragmentation, along with psychological distress appear to be important risk factors for weight gain.

In their several studies, Vgontzas’ groups investigated the association between subjective short sleep duration and incident obesity [49,50,51], demonstrating that this association is mediated by poor sleep and emotional stress. The results stratified by gender showed that the effect of emotional stress is stronger in women (p < 0.01 [50]) while complaints of poor sleep are strong predictors of incident obesity in men (p < 0.05 [51]). The role of depressive symptoms was investigated by Zimmerman et al. [22]. They studied the sleep onset/maintenance difficulties (SO/MD), finding a linear-by-linear association with BMI (χ 2 = 4.8, p < 0.03), abdominal obesity (χ 2 = 4.9, p < 0.03), depressive symptoms (χ 2 = 26.5, p < 0.01), and anxiety symptoms (χ 2 = 9.1, p < 0.01). The older participants showed the highest proportion of symptoms and higher proportion of moderate/severe SO/MD.

In an exercise-based randomized controlled trial with older adults (with health education as the comparison group), Buman and colleagues [52] studied mediators of the effect of exercise on improved objective sleep quality. Decreased depressive symptoms (95% CI − 0.75, 0.01), decreased BMI (95% CI − 1.08, 0.06), and increased physical function (95% CI 0.01, 0.72) mediated change in number of awakenings. Results suggest that sleep quality may benefit from moderate-intensity physical activity and that sleep-related benefits may have been conferred through reductions in depressive symptoms, reduction in BMI, or both. Similarly, Vaatainen et al. [21] found that the frequent awakenings were associated with a higher likelihood of both minor and moderate-to-severe depression as well as having more comorbidities. Participants with frequent nocturnal awakenings were more likely to be female and also had higher mean BMI. The prevalence of obesity was 25.4% in the group of frequent nocturnal awakenings and 17.5% (p < 0.01) in the infrequent nocturnal awakenings group.

Valentine and colleagues [53] found interesting sex differences in older age: more women reported being bothered by fatigue than men (65 vs. 40%, respectively, p < 0.05). They clearly found a sex difference in the relations among relative percent fat, C-reactive protein, physical activity in older men and women. In women, inflammation was a significant independent predictor of fatigue, explaining 6.6% of the variance, as were depression (6.3%), physical activity (5.8%), and percent fat (3.9%). The only independent predictor of fatigue in men was depression, accounting for 12% of the variance. Sleep quality was also associated with depression in men (r = 0.34, p = 0.02). The mechanisms responsible for the reported sex differences in fatigue are probably multi-factorial, including having a biological and psychosocial basis. A study by Cross et al. [39] already mentioned above, found a positive correlation between the number of naps per day and BMI, but no correlation with depressive symptoms. However, in the group who did take naps, depressive symptom severity was associated with longer duration of naps and a greater number of naps per day. Sleep–wake disturbance in older patients with depression has been shown to be related to neuropsychological functioning; the authors suggested the possibility that depressive, cognitive, and sleep-disturbance features may share common neurobiological underpinnings, which may aid in earlier identification of disease onset.

Beyond the depressive symptoms, Bidulescu et al. [54] demonstrated that perceived stress mediated the association between reduced sleep quality and obesity in a group of African Americans. They found that sleep quality and sleep duration were associated with categorized body mass index and the general sleep quality was associated with obesity among women (OR 1.08, 95% CI 1.03, 1.12) but not among men [OR (95% CI): 0.98 (0.89–1.09)]. Interestingly, there was an increased likelihood of obesity in the medium stress group of participants compared with low and high stress groups.

In a two-phase trial, Elder et al. [55] suggested that sleep time was a predictor of weight loss. They found that weight loss was significantly correlated with declines in stress (p = 0.048) and depression (p = 0.035). Unfortunately, the results are not stratified for age (overall age M = 55.3), but it is an interesting correlation of a possible inverse relation between sleep quality and obesity.

Physiological mediators and moderators

A possible mechanism for the poor sleep quality/obesity association is that chronic sleep restriction influences cardiovascular and metabolic health [56], with physiological consequences including impaired glucose tolerance and insulin sensitivity, elevated sympathetic tone, increased inflammation, and the increase of ghrelin and the decrease of leptin with the subsequent increase of hunger and appetite [57,58,59,60].

Few controlled experimental studies have addressed the question of the effects of sleep on body weight/composition to more specifically explicate mechanisms. The available literature suggests that sleep restriction increases food intake and total energy expenditure with inconsistent effects on integrated energy balance as operationalized by weight change. Laboratory studies have provided convincing evidence for a causal link between short sleep and poor sleep quality and adverse metabolic effects, although the mechanisms are not well understood (for a review 61, 62).

Ghrelin and Leptin are hormones which regulate energy balance and food intake and are increased during sleep [59].

The associations between self-reported sleep measures and Ghrelin or Leptin is not always consistent. In their longitudinal analyses, Littman et al. [15] followed a group of post-menopausal women during a physical activity intervention, hypothesizing that improved sleep might increase weight loss and attenuate exercise-induced increases in Ghrelin and decreases in Leptin. However, changes in sleep were not strongly associated with changes in weight, nor with Ghrelin and Leptin levels. However, looking at the effect of Ghrelin on sleep in a single-blind randomized experiment [63], the group tested with the Ghrelin injected compared to the placebo group showed in men a sleep pattern with significantly longer stage 2 sleep (S2S, p < 0.001), slow-wave sleep (SWS, p < 0.033) and non-REM sleep (p = 0.001), while stage 1 sleep (p = 0.012) and REM sleep (p = 0.049) were significantly reduced; these effects were not confirmed for women. In summary, Ghrelin had remarkably strong sleep-improving effects in elderly men, stronger than the effects found in younger people [64], where S2S was increased by 21% (young men: 10%) and SWS even by 33% (young men: 10%). The reason why no effect exists for older women is still unclear.

Jackowska et al. [65] used data from the English longitudinal study of aging (ELSA) to describe associations between biomarkers (C-reactive protein CRP, fibrinogen, dehydroepiandrosterone sulfate DHEAS, hemoglobin), sleep duration, and sleep disturbances. Men with longer sleep (> 8 h) had higher CRP levels compared with those sleeping 7–8 h (OR 1.50, CI 1.05, 2.14), and fibrinogen was higher among long sleepers (F (4,2845) = 5.007, p = 0.001). No relation with biomarkers and sleep duration was observed in women. With intermediate levels of disturbed sleep men reported elevated CRP levels (OR 1.29, CI 1.05, 1.59), and the same relationship was found when CRP was treated as a continuous variable (F (2,2912) = 7.420, p = 0.001). Always in men, more disturbed sleep was connected with lower levels of DHEAS (F (2,2899) = 4.126, p = 0.016), lower hemoglobin (F (2,2882) = 3.239, p = 0.039), and greater likelihood of anemia (OR 1.73, CI 1.13, 2.65). On the other hand, in women, more disturbed sleep was associated only with greater likelihood of anemia (OR 1.59, CI 1.02, 2.46).

In a more recent study [66] on a sample of post-menopausal women with high visceral abdominal adiposity, poorer overall sleep quality and greater visceral adiposity was linked to greater stress-induced increases in inflammatory activity (p = 0.17). Thus, sleep quality may be a modulator of stress responsiveness with consequences for putative inflammatory mechanisms related to chronic disease risk.

Recently, there has been interest in the idea that vitamin D could be important for sleep disorders [67]. Several cross-sectional studies have reported a high prevalence of vitamin D deficiency in obstructive sleep apnea syndrome (OSAS) and associated with abnormal glucose metabolism [68, 69]. It has been found in older men that low levels of total serum 25(OH)D are associated with poorer sleep including short sleep duration and lower sleep quality [70]. Low levels of vitamin D (< 20.3 ng/mL 25(OH)D) were found for 16.4% of the participants. Lower serum vitamin D levels were associated with higher odds of short sleep duration (< 5 h), OR for the highest: (≥ 40.06 ng/mL) vs. lowest (< 20.3 ng/mL) quartile of 25(OH)D, 2.15; CI 1.21, 3.79; p = 0.004. In addition, an independent association has been found also between serum 25(OH)D concentrations and depressive symptoms in an older population [71].

Poor sleep quality and low sleep duration have also been shown to negatively affect glucose regulation. Sleep curtailment alters energy expenditure, weight regulation, and inflammatory cytokine levels. The effects of sleep restriction are observed in healthy individuals with varying age and gender and in patients with medical conditions. The effect of sleep on the progression of many other medical conditions and their treatment is unknown. Despite the convincing evidence for the deleterious effects of decreased sleep quality and quantity, there is a paucity of research performed to test sleep extension or sleep improvement (and even poorer about older people) as a therapeutic approach to improve metabolic health in individuals who have acquired or are at risk for developing obesity.

Discussion

Societies in the developed world are witnessing steady increases in obesity and higher body mass index, including among older adults aged 60 and above, whose prevalence of obesity is estimated to be 37% [72]. Patterns of decline in sleep quantity and quality have also been increasing; in the US, the prevalence of habitual short sleep (< 6 h per night) is 18% or 53 million short sleepers who may be at risk of associated obesity [73]. While industrialization has brought greater efficiencies in production of goods and services, substantially fewer people today experience an agricultural way of life than their forebears, and this has resulted in more sedentary lifestyles and a sleep–wake cycle that is no longer tied to the period of daylight in each 24-h day [29]. Modern technological devices provide an ever-increasing opportunity to engage in activities other than sleep. Furthermore, what sleep is obtained is often fragmented and of poor quality due to a much greater range of life stressors in modern society (e.g., crime, terrorism, economic uncertainties, and unemployment).

Habitually receiving a good night’s sleep is important for both psychological and physiological well-being. Inadequate sleep can increase risk for depression [74], while good sleep supports positive health-related behaviors such as physical activity, which can be beneficial to maintenance of a healthy weight and overall good physical health. In this review, we have examined the evidence to date linking good sleep quality and quantity to physical health, with a focus on links to obesity.

Short duration [35], long duration [36], a U-shape distribution with both short and long sleep durations [37], as well as an increase in number of night-time awakenings [38], have all been linked to weight gain and/or risk for development of obesity, in longitudinal studies. Clearly, more studies with prospective designs and longer follow-up are needed.

Considering the evidence from the body of cross-sectional studies, sleep duration, as the most commonly used indicator of sleep quality, also reveals mixed findings. While some studies find no association at all linking sleep duration [28, 29, 31, 33, 34] and sleep quality [30] with measures of obesity, the majority of studies do find significant links. Again, both short [1, 12,13,14] and long [15, 16] sleep durations have been associated with obesity or higher BMI. Echoing results from longitudinal studies, a U-shape distribution suggesting both short (e.g., < 5 or 6 h/night) and long (e.g., > 7 or 8 h/night) sleep durations are also linked to greater obesity risk [17,18,19, 37]. In all but one [31] of the above-mentioned studies that failed to report a U-shape distribution, a non-linear association was actually not tested. Still, evidence is mounting that habitually getting too little or too much sleep can be detrimental.

It is of interest to note that it is not just sleep quantity that is linked to obesity risk. Problems with sleep initiation [22,23,24], nocturnal awakening [21, 22], and daytime sleepiness [25,26,27, 39], as self-report measures, have also been consistently associated with higher BMI. This suggests the importance of establishing healthier daily routines that will promote better sleep initiation and quality [75, 76], and consequently may help to lower risk of obesity.

While studies using self-report measures are helpful, studies using objective measures tend to increase internal validity of findings; however, there is variance in the accuracy of various objective methodologies. In a study of older adults, Sivertsen et al. [77] demonstrated that actigraphy has high sensitivity to detect sleep (95%), but performs less well for detecting wakefulness (sensitivity of 36%) and further, relative to polysomnography, actigraphy underestimates sleep-onset latency and over estimates total sleep time and sleep efficiency. Two studies in this review used polysomnography revealing that percentage of time in slow-wave sleep was predictive [45], and percentage of time in REM sleep was not predictive [41] of BMI. The remaining studies with objective data used actigraphy and while imperfect and ranging over a variety of different aspects to sleep quality, it is nevertheless of value to see a consistent pattern of findings with the actigraphy measures as regards links to obesity. These metrics have included percentage of time in impaired sleep–wake cycle [45], short sleep duration [43], and greater sleep variability [44, 47] as all being linked to higher BMI and/or obesity.

Of course, causality cannot be discerned with cross-sectional studies, and therefore, the direction of effect is still inconclusive. With ever-increasing body mass and obesity, individuals may become more prone to excessive or disturbed/shortened sleep due to problems such as disrupted breathing, snoring, sleep apnea, and restless leg syndrome [8, 78]. Conversely, persons who get too much or too little sleep or who experience regular sleep disturbances may experience more daytime fatigue, depressive symptoms, and, therefore, reduced motivation for physical activity (too tired so more sedentary life) or healthy eating patterns (awake longer so more opportunities to eat/snack) [43]. Furthermore, inadequate sleep curtails the brain’s ability to accomplish clearance of beta amyloid proteins typically experienced during restful sleep, as demonstrated by studies showing greater β-amyloid burden via PET imaging among persons with shorter sleep duration and lower sleep quality [79]. Lack of clearance of β-amyloid from the brain can affect cognitive functioning and may also indirectly affect health-related behaviors, with the resulting unhealthy lifestyle placing the individual at higher risk for ever increasing body mass.

Conclusions

In summary, the association between sleep and obesity has been observed in elderly populations, with some evidence of key psychological and physiological factors involved in underlying mechanisms including perceived stress and depression. Emotion regulation may also moderate the relationship between poor sleep and obesity as in the example of disinhibited eating behavior in some individuals when experiencing negative emotions, resulting in increased caloric intake leading to obesity [80]. Foley et al. [81] suggested that the sleep complaints common in older adults are often secondary to their comorbidities and not to aging per se. These types of studies may be useful in promoting sleep awareness among health professionals and among older adults, especially those with heart disease, depression, chronic bodily pain, or major comorbidity. In general, the influence of psychological factors on the sleep–obesity association reported in this review is similar to that found with younger populations. However, the influence seems to be not as strong among older adults potentially due to their higher comorbidity, but it should also be noted that there are insufficient data on older adults published to date to draw definitive conclusions.

More studies using objective measurement are needed, and in addition, studies employing controlled experimental designs would help further explicate this association. Without such studies, it is difficult to establish the direction of the causal relations. A cautionary note is that the effects of sleep deprivation in contrived experimental settings may not translate well to real-life experience; in addition, studies may need to examine both sleep quantity as well as quality as both may be important [82]. Insufficient sleep and unresolved stress may both be associated with weight gain and related eating patterns, although it is difficult to isolate one factor from another, as well as their cause–effect relationships. Important questions remain concerning the effect of poor sleep on obesity risk in older adults, such as: how can psychological factors affect sleep quality and lead to unhealthy eating patterns? Consequently, we suggest that future studies of older adults more closely examine the mechanisms, whereby psychological distress, disordered eating behavior, and medical comorbidities affect the sleep and obesity relationship, to more fully explain underlying processes. It will be important to clinicians to address comorbidities like depression and stress, and also with older patients to treat sleep dysfunction and obesity, and to avoid the causal circularity that seems to link these conditions.