Associations between sleep and body composition in older women and the potential role of physical function

Abstract

The relationship between sleep and adiposity in older women remains unclear partly due to the reliance of body mass index as a measure of adiposity. The purpose of this study was to investigate associations between objectively measured sleep characteristics and body composition measured by dual energy X-ray absorptiometry in older women. A secondary purpose was to examine if physical function mediates this relationship. Non-obese older women (ages 60–75 years, n = 102) were included in the study. Total sleep time (TST), time in bed (TIB), sleep efficiency (SE), and wake after sleep onset (WASO) were determined by actigraphy. A battery of tests was used to assess physical function. With adjustment for age, there was a negative association between TST and TIB with lean mass. Both grip strength and dominant leg extension were associated with TST, TIB, and lean mass; the associations between TST and TIB with lean mass were lost after adjusting for grip strength or leg extension strength. Additionally, SE was negatively associated with total, gynoid, and trunk lean mass, and there was a positive association between TST and percent trunk fat, and WASO and gynoid lean mass, with age adjusted. Sleep characteristics, TST, TIB, SE, and WASO, were associated body composition measures in this sample of older women. The relationship between TST and TIB with body composition was mediated, in part, by grip strength and leg extension strength.

Introduction

An increasing amount of research has examined associations between sleep duration and adiposity in adults. However, studies have found mixed results between body mass index (BMI) and sleep duration in older adults [1–3]. One reason for the inconsistent findings may be related to the assessment of sleep, i.e., self-report or objective measures. Compared to younger adults, older adults have shorter total sleep time (TST), worse sleep efficiency (SE), and greater wake time after sleep onset (WASO) [4–6]. These age-related changes in sleep patterns may increase the discrepancy between self-reported and objectively determined sleep.

Another important issue is the reliance on BMI as a measure of adiposity in previous studies. This is particularly an issue in postmenopausal women. A study done by Banack et al. [7] found that postmenopausal women classified as obese by percent body fat were misclassified as non-obese by BMI criteria. Thus, studies using a criterion method to objectively measure sleep and body composition are important to determine their associations in older adults.

Furthermore, sleep allows for physiological recovery, and long sleep duration has been hypothesized to reflect sleep need [8, 9]. In older adults, poor and long sleep may be a precursor or symptom of underlying conditions [4, 10–12]. Self-reported sleeping ≥ 8 h was associated with higher mortality than those who slept 7 h, while sleeping ≤ 6 h was not [12]. Physical function and cardiorespiratory fitness (CRF) are associated with morbidity and mortality [12–16]. Thus, women with lower physical function may sleep longer. On the other hand, aging-associated changes in body composition may put older women at greater risk for physical function decline [17, 18]. Therefore, physical function may mediate the observed association between body composition and sleep.

The purpose of this study was to investigate associations between objectively measured sleep characteristics [time in bed (TIB), TST, WASO, and SE] and body composition measured by dual energy X-ray absorptiometry (DXA) in older women. The secondary purpose of the study was to examine if physical function and CRF mediate the relationship between sleep characteristics and body composition.

Materials and methods

Data were from the Women’s Energy Expenditure in Walking Programs (WEWALK) [19] and the Calorie Expenditure and Exercise (CEE) studies [20]. The two studies used the same inclusion and exclusion criteria. Briefly, participants were healthy, non-obese (BMI 18.5–30 kg/m²), older women (age 60–75 years), physically inactive (did not purposely exercise more than 20 min three times per week), and weight stable for 3 months (change <  ± 3%). Participants were excluded if they reported major cardiovascular, metabolic, or respiratory diseases or severe sleep problems. Participants with signs of cognitive dysfunction (Mini-Mental State Examination < 24) or depression (Center for Epidemiologic Studies Depression Scale > 16) were excluded. The CEE study (clinicaltrials.gov registration: NCT00988299) protocol was approved by the Institutional Review Board (IRB) of Washington University School of Medicine, St. Louis, MO. The WEWALK study (clinicaltrials.gov registration: NCT01722136) protocol was approved by the IRB of the University of South Carolina, Columbia, SC. All participants signed an informed consent form prior to participation.

Body composition

Height and weight were measured twice using a stadiometer and digital scale, respectively. Averages were used to calculate BMI. A DXA scan was used to measure whole-body fat mass, lean mass, and percent body fat (WEWALK: enCORE, GE Healthcare model 8743, Waukesha, WI; CEE: Hologic QDR 1000/w, Waltham, MA, USA). In the WEWALK study, the DXA scan also provided fat mass, lean mass and percent body fat for the trunk, android (area between the ribs and the iliac crest), and gynoid (area around the hips) regions [21]. Visceral adipose tissue (VAT) within the android region was estimated.

Sleep characteristics

Participants wore an accelerometer (GT3X + ; Actigraph, Pensacola, FL, USA) on their non-dominant wrist for up to 14 days. They reported bedtimes and wake times on a daily sleep log. The manufacturer’s software (ActiLife version 6.11.2) was used to analyze the actigraphy data in 60-s epochs. Sleep log times were manually entered into the program. A standardized approach was used to define TIB by estimating missing or adjusting inaccurate bed/awake times using a hierarchical ranking of inputs (i.e., sleep logs, light intensity, and activity counts). Sleep and awake status were determined by the Cole–Kripke algorithm for each TIB interval [22]. The software provides estimates for WASO (total time that the algorithm scored as “awake” after sleep onset) and TST. SE was calculated as TST divided by TIB. Three nights of wear were considered valid for analysis. In the CEE study, the average was 6.1 days, ranging from 4 to 8 days. In WEWALK, the average was 13.3 days, with only 1 participant having < 7 days of valid data.

Cardiorespiratory fitness (CRF)

A graded treadmill exercise test was used to measure peak oxygen consumption ($\dot{\text{V}}$O_2peak) which served as an index of CRF. Participants walked at a constant speed while grade increased every two minutes until volitional exhaustion. A metabolic measurement system (TrueOne 2400, Parvo Medics, Salt Lake City, UT, USA) was used for gas analysis. For a $\dot{\text{V}}$O_2peak to be considered valid, two of the following criteria needed to be met: plateau in $\dot{\text{V}}$O₂ (a change < 2 mL/kg/min) with increasing work rate, highest heart rate > 90% of age-predicted maximum heart rate, respiratory exchange ratio ≥ 1.10, and a rate of perceived exertion ≥ 17 on the Borg exertion scale.

Physical activity and physical function

Counts per minute (CPM) were calculated from accelerometer data of the entire wear time, including nighttime sleep, using the manufacturer’s software, which is a measure of total physical activity, integrating information from all 3 axis. Physical function was only measured in the WEWALK study and included 4-m walk time, grip strength, and isokinetic leg strength. To determine walk time, participants walked 4 m at their maximal pace. The average time of three trials was used for analysis. Grip strength was measured using an adjustable digital dynamometer (Jamar Plus + Sammons Preston, Warrenville, IL, USA) with participants seated and elbow flexed at 90 degrees. The average of three trials for each hand was used for analysis. Maximal leg strength was measured using an isokinetic dynamometer (Biodex System 3 Pro, Biodex Medical Systems Inc., Shirley, NY, USA). Participants were seated with their hips and knees at 90° of flexion and secured to the chair with a harness. The knee axis of rotation was aligned with the dynamometer shaft. The knee attachment was secured with a strap above the ankle. Bilateral isokinetic (concentric/concentric) flexion and extension of the knee were performed 6 times. Test speed was set to 60°/s and the participant was instructed to push as hard as possible to flex or extend the knee being tested. Participants performed a test trial before starting the test and 30 s of rest was given between trials. The average of the two highest reproducible values (within 10% of each other) was used for analysis.

Statistical analysis

SAS software, version 9.4 (SAS Institute, Cary, NC, USA) was used for analysis. Means and standard deviations (SD) and medians (interquartile range, IQR) were calculated for each variable. General linear models were performed to assess associations between sleep characteristics and body composition. These models were adjusted for site (when data from both sites were used) in the initial model (model 1), and site and age in model 2. The associations that were identified to be significant were further evaluated with adjustment for physical function variables or CRF to determine their potential mediating role (model 3). A physical function variable or CRF was selected as a potential mediator to be adjusted in model 3 if it was significantly associated with both sleep characteristics and body composition. A p < 0.05 was considered statistically significant.

Results

Participant characteristics

The current analysis included 14 participants in CEE and 88 participants in WEWALK who had DXA scan and sleep data. In the CEE study, the DXA scan did not determine segmental composition and physical function tests were not performed. Of the WEWALK participants, 28, 31, 29, and 27 missed grip strength, leg extension, leg flexion, and 4-m walk, respectively. Participants were older (mean age: 65.5 ± 4.2 years), non-obese, and mostly white (84.1%) women. Most slept 7–9 h (57%), 36% slept 6–7 h, 6% slept < 6 h, and 1% slept > 9 h per night. The mean TST, TIB, SE, and WASO were 429.0 ± 48.4 (range 318.6–541.4) min, 477.7 ± 52.0 (range 346.5–616.3) min, 90% (range 76–97%), and 43.6 ± 20.2 (range 8.7–128.3) min, respectively. Sleep characteristics were similar between the two studies (p > 0.6), except for WASO (CEE = 54.2 ± 36.4 min vs. WEWALK 41.7 ± 16.9 min; p < 0.001). Few participants (< 10%) reported taking daytime naps, therefore naps were not specifically examined.

Descriptive statistics for body composition, physical function, and CRF are in Tables 1 and 2, respectively. Measurements were similar between the two studies, except for CRF ($\dot{\text{V}}$O_2peak = 22.5 ± 4.01 (CEE) versus 20.2 ± 3.6 (WEWALK) ml/kg/min; p = 0.008).

Table 1.

Whole and segmental body composition descriptive statistics

Body composition	N	Mean ± SD	Median (IQR)
Weight (kg)	102	67.8 ± 9.3	66.1 (61.4–76.4)
BMI (kg/m2)	102	25.6 ± 3.3	25.5 (23.2–28.0)
Fat mass (kg)	102	26.6 ± 6.5	26.4 (22.9–30.6)
Lean mass (kg)	102	39.4 ± 4.32	38.6 (36.0–42.2)
% Body fat	102	38.5 ± 5.7	38.7 (36.1–43.3)
*Gynoid fat mass (g)	88	4813 ± 1187	4678 (4002–5668)
*Gynoid lean mass (g)	88	5916 ± 701	5895 (5427–6373)
*% Gynoid fat	88	43.5 ± 5.5	43.9 (39.5–47.4)
*Android fat mass (g)	88	2138 ± 756	2131 (1711–2644)
*Android lean mass (g)	88	2810 ± 347	2820 (2589–3041)
*% Android fat	88	41.5 ± 9.5	44.1 (37.5–47.9)
*Trunk fat mass (g)	88	13,015 ± 3903	13,083 (11,217–15,820)
*Trunk lean mass (g)	88	18,396 ± 2043	18,465 (17,030–19,771)
*% Trunk fat	88	40.0 ± 7.7	41.6 (36.3–45.3)
*VAT Mass (g)	88	868 ± 485	797 (512–1155)
*VAT Volume (cm3)	88	921 ± 514	845 (543–1224)

BMI, body mass index; VAT, visceral adipose tissue; SD, standard deviation; IQR, interquartile range

*Sample from WEWALK only

Table 2.

Physical function and cardiorespiratory fitness descriptive statistics

Test variable	n	Mean ± SD	Median (IQR)
Dominant leg extension (N·m)	57	81.3 ± 26.7	77.2 (62.2–102.8)
Nondominant leg extension (N·m)	57	70.1 ± 21.9	69.7 (51.4–81.3)
Dominant leg flexion (N·m)	59	39.9 ± 13.1	42.8 (27.9–50.4)
Nondominant leg flexion (N·m)	59	34.1 ± 13.1	35.4 (24.1–44.3)
Dominant hand grip (kg)	60	25.9 ± 5.2	24.9 (22.4–30.2)
Nondominant hand grip (kg)	60	22.8 ± 5.9	22.6 (18.5–26.8)
4-meter walk (s)	61	4.0 ± 0.7	3.9 (3.4–4.5)
*Counts per minute	102	1577.5 ± 425.8	1531.8 (1286.1–1874.0)
*V̇O2peak (ml/kg/min)	102	20.3 ± 3.9	19.9 (17.6–22.4)

$\dot{\text{V}}$ O_2peak, Peak oxygen consumption; SD, standard deviation; IQR, interquartile range

*Sample from WEWALK and CEE

Associations between sleep characteristics and body composition

Associations between sleep characteristics and whole-body composition are displayed in Table 3. In model 1, TIB and TST were negatively associated with lean mass (p = 0.015 and 0.0021, respectively). For every 10-min longer TIB or TST, lean mass was less by 204.7 or 275.2 g, respectively. These associations remained after adjusting for age (p < 0.03 for both, model 2). Body weight was associated with TST (p = 0.0433) only in model 1. SE was associated with lean mass when age was adjusted for (p = 0.033, model 2). Fat mass, BMI, and percent body fat were not significantly associated with TST, TIB, SE, or WASO.

Table 3.

Associations between sleep characteristics and whole-body composition

		Model 1								Model 2
		TIB (min)		TST (min)		SE (%)		WASO (min)		TIB (min)		TST (min)		SE (%)		WASO (min)
	n	β (95% CI)	r	β (95% CI)	r	β (95% CI)	r	β (95% CI)	r	β (95% CI)	r	β (95% CI)	r	β (95% CI)	r	β (95% CI)	r
Weight (kg)	102	− 0.026 (− 0.061, 0.0093)	− 0.15	− 0.038 (− 0.076, − 0.0012)**	− 0.20**	− 0.41 (− 0.89, 0.068)	− 0.17	0.058 (− 0.035, 0.15)	0.12	− 0.022 (− 0.058, 0.014)	− 0.12	− 0.035 (− 0.073, 0.0023)	− 0.19	− 0.45 (− 0.93, 0.032)	− 0.19	0.066 (− 0.028, 0.16)	0.14
BMI (kg/m.2)	102	0.0022 (− 0.010, 0.015)	0.035	− 0.0011 (− 0.014, 0.012)	− 0.017	− 0.12 (− 0.29, − 0.053)	− 0.14	0.023 (− 0.0099, 0.056)	0.13	0.0033 (− 0.0095, 0.016)	0.052	− 0.00026 (− 0.013, 0.013)	− 0.0039	− 0.12 (− 0.29, 0.048)	− 0.14	0.025 (− 0.0085, 0.058)	0.15
Fat Mass (g)	102	− 2.24 (− 26.78, 22.31)	− 0.018	− 6.65 (− 32.89, 19.55)	− 0.051	− 165.57 (− 499.20, 168.065)	− 0.098	27.36 (− 37.25, 91.97)	0.084	− 0.89 (− 25.91, 24.19)	− 0.0069	− 5.42 (− 32.06, 21.21)	− 0.041	− 172.22 (− 510.01, 165.57)	− 0.10	29.34 (− 36.17, 94.86)	0.090
Lean Mass (g)	102	− 20.47 (− 36.96, − 3.98)**	− 0.24**	− 27.52 (− 44.84, − 10.22)*	− 0.30*	− 211.81 (− 439.97, 16.35)	− 0.18	25.51 (− 18.050, 71.076)	0.11	− 18.68 (− 35.13, − 2.22)**	− 0.22**	− 26.39 (− 43.54, − 9.23)*	− 0.29*	− 244.26 (− 467.69, − 20.83)**	− 0.22**	33.20 (− 10.62, 77.02)	0.15
% Body Fat	102	0.00015 (− 0.000057, 0.00037)	0.14	0.00016 (− 0.0000062, 0.00039)	0.14	0.000025 (− 0.0029, 0.0029)	− 0.0018	− 0.000088 (− 0.00047, 0.00065)	0.031	0.00015 (− 0.000064, 0.00037)	0.14	0.00016 (− 0.000066, 0.00039)	0.14	0.00012 (− 0.0028, 0.0031)	0.0082	0.000062 (− 0.00051, 0.00063)	0.022

Data presented as regression coefficient (95% CI) and Pearson’s correlations (r). Body composition variables were the dependent variable in models. Model 1 is adjusted for site. Model 2 is adjusted for site and age. Significant findings are in bold. *p ≤ 0.01; **p < 0.05.

TIB, time in bed; TST, total sleep time; WASO, wake after sleep onset; CI, confidence interval

Associations between sleep characteristics and segmental body composition were examined using WEWALK study data only (Table 4). In model 1, gynoid lean mass was negatively associated with TST, TIB, and SE (p = 0.004, p = 0.025, p = 0.020 respectively). For every 10-min longer TST and TIB, gynoid lean mass was less by 43.3 and 32.6 g, respectively. For every 1% increase in SE, gynoid lean mass decreased by 51.5 g. Trunk lean mass was negatively associated with TST and SE (p = 0.025, p = 0.0266, respectively). When adjusting for age (model 2), the associations of TIB and TST with trunk lean mass and gynoid lean mass were lost. The association between SE with trunk lean mass and gynoid lean mass remained. There was a positive association between TST and trunk percent fat (p = 0.042) and a positive association between gynoid lean mass and WASO (p = 0.039). No other significant associations existed between segmental body composition variables and sleep characteristics.

Table 4.

Associations between sleep characteristics and segmental body composition

		Model 1								Model 2
		TIB (min)		TST (min)		SE (%)		WASO (min)		TIB (min)		TST (min)		SE (%)		WASO (min)
	n	β (95% CI)	r	β (95% CI)	r	β (95% CI)	r	β (95% CI)	r	β (95% CI)	r	β (95% CI)	r	β (95% CI)	r	β (95% CI)	r
Gynoid Fat Mass (g)	88	− 0.012 (− 4.80, 4.94)	− 0.00053	− 1.18 (− 6.38, 4.01)	− 0.048	− 57.88 (− 132.44, 16.67)	− 0.16	10.64 (− 4.25, 25.54)	0.15	2.24 (− 3.59, 8.084)	0.092	1.65 (− 4.47, 7.78)	0.065	− 22.39 (− 106.38, 61.59)	− 0.064	6.61 (− 9.88, 23.11)	0.096
Gynoid Lean Mass (g)	88	− 3.25 (− 6.10, − 0.41)	− 0.23	− 4.33 (− 7.27, − 1.40)	− 0.30	− 51.54 (− 94.79, − 8.29)	− 0.25	6.64 (− 2.13, 15.43)	0.16	− 1.53 (− 4.72, 1.66)	− 0.11	− 2.90 (− 6.19, 0.39)	− 0.20	− 56.14 (− 100.25, − 12.026)	− 0.29	9.31 (0.50, 18.12)	0.24
% Gynoid Fat	88	0.00019 (− 0.000035, 0.00042)	0.17	0.00018 (− 0.000056, 0.00042)	0.16	− 0.00047 (− 0.0040, 0.0031)	− 0.029	0.00025 (− 0.00044, 0.00096)	0.077	0.00023 (− 0.000040, 0.00051)	0.20	0.00026 (− 0.000028, 0.00055)	0.21	0.0011 (− 0.0029, 0.0052)	0.065	0.0000074 (− 0.00079, 0.00081)	0.0022
Android Fat Mass (g)	88	1.22 (− 1.91, 4.36)	0.083	0.93 (− 2.37, 4.24)	0.060	− 12.97 (− 61.01, 35.08)	− 0.058	3.76 (− 5.79, 13.33)	0.084	2.81 (− 1.12, 6.74)	0.17	2.67 (− 1.44, 6.80)	0.15	− 2.19 (− 59.42, 55.04)	− 0.0092	3.17 (− 8.07, 14.41)	0.068
Android Lean Mass (g)	88	− 0.81 (− 2.25, 0.62)	− 0.12	− 1.17 (− 2.68, 0.32)	− 0.16	− 17.49 (− 39.27, 4.29)	− 0.17	2.59 (− 1.77, 6.97)	0.12	− 0.13 (− 1.82, 1.54)	− 0.020	− 0.57 (− 2.33, 1.18)	− 0.078	− 17.75 (− 41.55, 6.065)	− 0.18	3.30 (− 1.39, 8.00)	0.16
% Android Fat	88	0.00032 (− 0.000072, 0.00071)	0.17	0.00032 (− 0.000090, 0.00072)	0.16	0.00018 (− 0.0059, 0.0062)	0.0065	0.00026 (− 0.00094, 0.0014)	0.046	0.00047 (− 0.000029, 0.00097)	0.22	0.00049 (− 0.000033, 0.0010)	0.22	0.0012 (− 0.0062, 0.0086)	0.038	0.00018 (− 0.0012, 0.0016)	0.031
Trunk Fat Mass (g)	88	5.37 (− 10.87, 21.62)	0.070	4.08 (− 13.014, 21.17)	0.051	− 61.29 (− 309.39, 186.82)	− 0.13	16.72 (− 32.69, 66.14)	0.50	14.93 (− 5.11, 34.98)	0.17	14.56 (− 6.46, 35.59)	0.16	− 0.77 (− 293.022, 291.47)	− 0.075	14.20 (− 43.24, 71.66)	0.059
Trunk Lean Mass (g)	88	− 7.01 (− 15.40, 1.37)	− 0.17	− 10.00 (− 18.70, − 1.30)	− 0.23	− 143.43 (− 269.80, − 17.068)	− 0.24	20.12 (− 5.45, 45.7)	0.12	− 0.71 (− 10.41, 8.99)	− 0.017	− 4.90 (− 15.00, 5.18)	− 0.11	− 169.21 (− 302.33, − 36.082)	− 0.29	31.07 (4.69, 57.45)	0.27
% Trunk Fat	88	0.00028 (− 0.000031, 0.00060)	0.18	0.00030 (− 0.000030, 0.00063)	0.19	0.00082 (− 0.0041, 0.0058)	0.036	0.00011 (− 0.00087, 0.0011)	0.024	0.00040 (− 0.0000082, 0.00080)	0.23	0.00044 (0.000016, 0.00086)	0.24	0.0019 (− 0.0040, 0.0079)	0.080	− 0.000028 (− 0.0012, 0.0011)	− 0.0058
VAT mass (g)	88	0.51 (− 1.50, 2.53)	0.054	0.68 (− 1.44, 2.80)	0.068	8.63 (− 22.19, 39.45)	0.059	− 0.96 (− 7.11, 5.19)	− 0.033	1.28 (− 1.22, 3.80)	0.12	1.53 (− 1.094, 4.15)	0.13	11.89 (− 24.35, 48.13)	0.079	− 0.92 (− 8.078, 6.23)	− 0.031
VAT volume (cm3)	88	0.54 (− 1.60, 2.68)	0.054	0.72 (− 1.52, 2.97)	0.068	9.14 (− 23.53, 41.81)	0.059	− 1.020 (− 7.54, 5.50)	− 0.33	1.36 (− 1.30, 4.02)	0.12	1.62 (− 1.16, 4.40)	0.13	12.59 (− 25.81, 51.0031)	0.079	− 0.97 (− 8.56, 6.60)	− 0.031

Data presented as regression coefficient (95% CI) and Pearson’s correlations (r). Body composition variables were dependent variable in models. Data are from the WeWalk study. Model 1 is unadjusted. Model 2 is adjusted for age. Significant findings with p < 0.05 are bolded

TIB, time in bed; TST, total sleep time; WASO, wake after sleep onset; VAT, visceral adipose tissue; CI, confidence interval

Mediation role of physical function and CRF in associations between sleep characteristics and body composition

To determine the mediating role of physical function and CRF in associations between sleep characteristics and body composition, the associations of physical function and CRF with sleep characteristics were first examined. Significant associations are displayed in Table 5. In model 1, TST was negatively associated with extension of the dominant and non-dominant leg, dominant and non-dominant grip strength, and CPM (p value range: 0.001–0.026). TIB was negatively associated with dominant leg extension, grip strength, and CPM (p value range: 0.004–0.024). Most of the significant associations remained after adjusting for age, except associations between TIB with dominant leg extension and TST with non-dominant leg extension were lost. Additionally, there was a negative association between SE and dominant leg extension (p = 0.014). Knee flexion, 4-m walk time, and $\dot{\text{V}}$O_2peak were not significantly associated with TST, TIB, or SE. There were no significant associations between WASO and physical function or $\dot{\text{V}}$O_2peak.

Table 5.

Significant Associations between physical function with sleep characteristics and body composition

	Dominant leg extension (N·m) n = 57		Nondominant leg extension (N·m) n = 57		Dominant grip (kg) n = 60		Nondominant grip (kg) n = 60		CPM n = 102
	β (95%CI)	r	β (95%CI)	r	β (95%CI)	r	β (95%CI)	r	β (95%CI)	r
Model 1
TST (min)	− 0.20 (− 0.34, − 0.055)	− 0.35	− 0.13 (− 0.25, − 0.016)	− 0.29	− 0.046 (− 0.076, − 0.015)	− 0.37	− 0.049 (− 0.082, − 0.016)	− 0.44	− 2.84 (− 4.49, − 1.18)	− 0.32
TIB (min)	− 0.17 (− 0.32, − 0.023)	− 0.29	––	––	− 0.044 (− 0.075, − 0.013)	− 0.35	− 0.043 (− 0.077, − 0.0093)	− 0.35	− 2.35 (− 3.93, − 0.79)	− 0.29
SE (%)	––	––	––	––	––	––	––	––	––	––
Weight (kg)	0.18 (0.11, 0.26)	0.54	0.22 (0.12, 0.31)	0.53	0.72 (0.32, 1.12)	0.42	0.57 (0.20, 0.93)	0.39	––	––
Lean Mass (g)	0.0031 (0.0018, 0.0044)	0.55	0.0021 (0.0010,.0032)	0.47	0.00073 (0.00049, 0.00097)	0.63	0.00070 (0.00040, 0.00099)	0.53	––	––
Gynoid Lean Mass (g)	0.016 (0.0075, 0.024)	0.45	.012 (0.0047, 0.019)	0.41	0.0052 (0.0038,.0067)	0.69	0.0053 (0.0037, 0.0071)	0.63	0.13 (0.0065, 0.26)	0.22
Trunk Lean Mass (g)	0.0055 (0.0026, 0.0085)	0.45	0.0039 (0.0015, 0.0065)	0.39	0.0013 (0.00072, 0.0019)	0.50	0.0012 (0.00046, 0.0019)	0.39	0.047 (0.0033, 0.091)	0.22
Model 2
TST (min)	− 0.20 (− 0.38, − 0.023)	− 0.34	––	––	− 0.043 (− 0.073, − 0.14)	− 0.40	− 0.049 (− 0.082, − 0.016)	− 0.44	− 2.83 (− 4.48, − 1.17)	− 0.32
TIB (min)	––	––	––	––	− 0.040 (− 0.070, − 0.010)	− 0.32	− 0.043 (− 0.077, − 0.0093)	− 0.35	− 2.30 (− 3.88, − 0.73)	− 0.28
SE (%)	− 2.78 (− 4.97, − 0.60)	− 0.37	––	––	––	––	––	––	––	––
Weight (kg)	0.19 (0.11, 0.27)	0.59	0.22 (0.12, 0.32)	0.54	0.68 (0.25, 1.10)	0.45	0.53 (0.14, 0.91)	0.35	––	––
Lean Mass (g)	0.0028 (0.0014, 0.0041)	0.50	0.0018 (0.00068, 0.0030)	0.40	0.00070 (0.00045, 0.00095)	0.59	0.00064 (0.00033, 0.00094)	0.48	––	––
Gynoid Lean Mass (g)	0.015 (0.0051, 0.026)	0.69	0.0096 (0.00066, 0.019)	0.61	0.0051 (0.0036, 0.0067)	0.41	0.0051 (0.0032, 0.0070)	0.31	––	––
Trunk Lean Mass (g)	0.0049 (0.0013, 0.0087)	0.38	0.0033 (0.00065, 0.0059)	0.32	0.0012 (0.00055, 0.0018)	0.46	0.00094 (0.00021, 0.0017)	0.31	––	––

Data presented as regression coefficient (95% CI) and Pearson’s correlations (r). Physical function variables were dependent variables in models. Model 1 is unadjusted. Model 2 is adjusted for age. All models that used data from both studies are adjusted for site

p < 0.05 for all reported values

TIB, time in bed; TST, total sleep time; CPM, Counts Per Minute; CI, confidence interval

Next, the associations of physical function and CRF with body composition were examined. This was limited to those physical function variables and body composition measures found to be significantly associated with sleep characteristics (shown in Tables 3, 4, 5). In Model 1, body weight, lean mass, gynoid lean mass, and trunk lean mass were positively associated with dominant and non-dominant leg extension and grip strength (all p values ≤ 0.003; Table 5). All significant associations remained when age was adjusted (model 2). Gynoid and trunk lean mass were positively associated with CPM in model 1 (p values < 0.03), but the associations were lost in model 2.

Dominant and non-dominant leg extension and grip strength were therefore identified as potential mediators. These variables were individually adjusted in the models where significant associations between sleep characteristics and body composition were found. When adjusted for grip strength or leg extension, the associations between TST with weight, lean mass, gynoid lean mass, and trunk lean mass were lost (p value range: 0.097–0.987). The associations between TIB with lean mass and gynoid lean mass were also lost (p value range: 0.259–0.864). Dominant and non-dominant leg extension and grip strength all remained significantly associated (p < 0.05) with TST or TIB in these models. However, dominant leg extension did not mediate the relationship between SE with lean mass, or trunk or gynoid lean mass (p values < 0.05). Since WASO and percent trunk fat were not significantly associated with physical function, their association was not examined for mediation.

Discussion

In this sample of older women, several significant associations between body composition and sleep characteristics were found. Among those, the associations between whole-body lean mass with TST and TIB, and gynoid and trunk lean mass with SE remained after adjusting for age. When examining physical function, grip strength and leg extension were identified as mediators of the associations between sleep characteristics and body composition.

In studies using a criterion measure of body composition, the main findings are that short sleepers (< 5 h) and long sleepers (> 8 h) had greater fat mass or percent body fat compared to individuals with a sleep duration between 7 and 8 h [2, 23]. We only found a positive association between TST and trunk percent fat, and not with percent body fat. However, previous studies had different participant characteristics and a higher percentage of short and long sleepers compared to our sample. We also found a negative association with TST and lean mass. Research on lean mass and TST is limited. However, one study found self-reported long sleep duration (> 9 h) to be associated with a decrease in lean mass in adults, ages 18–70 years [24]. Our results indicate in older women with 93% sleeping 6–9 h, there may still be an association between sleep duration and lean mass [25].

Poor physical function has been associated with longer sleep duration in previous studies [26–28]. One study in postmenopausal women found a negative correlation between sleep duration with functional capacity and muscle mass [27]. The authors speculated that sleeping longer leaves less time to be physically active, leading to decreased physical function and muscle mass. Longitudinally, women who reported consistently sleeping greater than 7 h per night, had higher odds of having a balance problem at the end of 13-year follow-up than those reporting consistently sleeping less than 6 h per night [29]. In our study, leg extension and grip strength were found to be negatively associated with TIB and TST, in line with previous findings regarding sleep and physical function.

Sleep is a restorative process that is important for physical health [30]. Previous research indicates that poor sleep in older adults may be a precursor or symptom of underlying conditions, and hypothesized that long sleep reflects sleep need [4, 9–11]. Therefore, we hypothesize physical function is mediating the relationship between body composition and sleep duration in older adults. As defined by Baron and Kenny [31], for a mediation to be established, the independent variable (X; TST) must be associated with the dependent variable (Y; lean mass). X must be associated with the mediator variable (M; grip strength or leg extension strength). M must be significantly associated with Y after adjusting for X. Lastly, the effects of X on Y should decrease when including M in the model [32]. All criteria were met, indicating grip strength and leg extension strength play a mediating role in the relationship between TST and whole-body, gynoid, and trunk lean mass, and between TIB with gynoid lean mass. Interestingly, we did not find CPM, 4-m walk time, $\dot{\text{V}}$O_2peak, or leg flexion as mediators of the relationship between sleep and body composition. Four-meter walk time and $\dot{\text{V}}$O_2peak are indicative of whole-body function and are influenced by multiple body systems and their relationship with sleep may be more complicated. Grip strength and leg extension strength, however, are primarily determined by muscle function. Muscle strength declines differentially among muscle groups in older adults. In older women, leg flexion strength decreases at a faster rate than leg extension strength [33]. This may help explain why leg flexion was not a mediator, but leg extension was.

The main strength of the present study is having objectively measured sleep characteristics and DXA-measured body composition. Our findings revealed associations between sleep characteristics with whole body, trunk, and gynoid body composition. Using BMI alone would have missed these associations. Another strength was our population were mostly normal sleepers, with 57% sleeping 7–9 h a night with an average WASO of 43.66 ± 20.18 min. Therefore, our study showed associations between body composition and sleep characteristics even within mostly normal sleepers. However, this study also has limitations. The cross-sectional design does not allow us to determine if longer sleep precedes or follows decreases in lean mass and physical function. The study population was healthy, normal to overweight, older women, and thus not generalizable to other populations. Additionally, our sample size was small especially for segmental body composition and physical function measures. Lastly, CPM was calculated for the entire wear time regardless of activity being voluntary or involuntary, and we were not able to determine intensity of activity using the wrist-worn device. Despite the weaknesses, this study makes important contribution to our knowledge given the limited number of studies that objectively measured sleep, used criterion method to measure body composition, and examined physical function.

In conclusion, several associations between sleep characteristics and body composition were found. These relationships between TST and TIB with lean mass were mediated, at least in part, by leg extension and grip strength. These results would have been missed if only BMI was used, indicating a need for criterion methods of measuring body composition to be used in future research. More research is needed to examine possible mechanisms behind the relationship of sleep with body composition and physical function in older women.

Funding

This work was supported by the National Institutes of Health (NIH) under grants K99AG031297, R00AG031297, P30 DK056341 (Nutrition and Obesity research Center), UL1 RR024992 (Washington University School of Medicine Clinical Translational Science Award).

Declarations

Conflict of interest

The authors declare that they have no conflict of interest.

Ethics approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the Institutional Review Boards at the University of South Carolina and Washington University in St. Louis and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Informed consent

Informed consent was obtained from all individual participants in the study.