Keywords
body mass index - overweight - sleep - sleep quality - lifestyle
Introduction
Obesity remains one of the most persistent global epidemics, presenting significant
challenges in terms of prevention, control, and treatment. Projections from the World
Obesity Federation[1] indicate that the prevalence of overweight and obesity will rise from 38% of the
global population in 2020 to more than 50% by 2035. In Brazil, this trend is equally
concerning, with estimates suggesting that, by 2030, 68.1% of the population will
be overweight or obese.[2] This significant public health issue has paralleled a decline in both sleep duration
and quality, alongside a growing prevalence of sleep disorders. Approximately one-third
of the global population is estimated to suffer from insufficient sleep.[3] This trend is expected to worsen in the coming years, driven by the increasing role
of technology and the pressures of a fast-paced, 24-hour modern lifestyle.[4]
Research increasingly highlights the interconnection between the rising prevalence
of excess weight and sleep-related problems, suggesting that these trends are not
merely coincidental.[5] While a U-shaped association between sleep duration and body weight has been observed,[6] with both short and long sleep linked to increased body weight, insufficient sleep
remains the primary concern in modern society. Beyond its role in obesity,[6]
[7] insufficient sleep has also been identified[8] as a significant risk factor for several other non-communicable chronic diseases,
including cardiovascular diseases, osteoporosis, stroke, and type-2 diabetes.
Although the precise mechanisms linking short sleep duration to obesity are not fully
understood, three key pathways have been proposed[9] to explain how insufficient sleep disrupts energy balance and contributes to weight
gain: first, sleep restriction alters metabolic hormones, increasing ghrelin levels
while reducing those of leptin, which together stimulate appetite; second, inadequate
sleep impairs glucose metabolism, further intensifying hunger; lastly, the fatigue
resulting from insufficient sleep reduces overall energy expenditure, contributing
to a more sedentary lifestyle.
Building on this understanding of sleep's role in metabolic health, recent research[10] highlights notable sex differences in sleep quality, with female subjects reporting
poorer sleep and more frequent disturbances compared with male subjects. These disparities
are influenced by a range of social, biological, and hormonal factors, particularly
during key life stages such as menstruation, pregnancy, and menopause, which can further
disrupt respiration, mood, and body temperature.[11] For instance, Fan et al.[12] found that female subjects with short sleep duration had a 62% higher likelihood
of developing general obesity and a 22% higher likelihood of developing visceral obesity,
whereas no such association was observed in male subjects. These findings, alongside
those of other studies,[7]
[13]
[14]
[15] suggest that the relationship between sleep and body composition may be more pronounced
in female than in male subjects.
Despite these findings, research on sex differences regarding the relationship between
nutritional status and sleep remains limited, emphasizing the need for further investigation.
To our knowledge, no studies have examined this comparison in Latin or South America.
To address this gap, the primary objective of the current study is to examine the
associations of sleep quality and duration with the body mass index (BMI) and excess
weight among male and female participants, hypothesizing that these associations will
be stronger in female than in male subjects. For the identified associations, we will
further investigate potential mediators, including a range of lifestyle and health-related
factors, which may influence these relationships.
Materials and Methods
Participants
The present cross-sectional study used data from the third cycle of the Sonar-Brazil
Survey, which was conducted between May 2023 and May 2024[38]. The survey is virtual, exploratory, and population-based, aiming to investigate
chronobiological aspects related to sleep, food, and nutrition among Brazilian adults.
The inclusion criteria for Sonar-Brazil were: being Brazilian, residing in Brazil,
being aged between 18 and 65 years, and not being pregnant. Much effort was made to
ensure the sample would be representative of the general population. The survey link
was emailed to all public and private higher education institutions registered with
the Brazilian Ministry of Education, as well as to regional nutritionist, dental,
and medical councils; health units; electronic, printed, and televised newspapers;
sleep and nutrition congresses; and scientific events. The survey was also widely
promoted on social media platforms and web sites, with a request for the link to be
shared with eligible contacts. Dedicated social media pages were established to enhance
the survey's visibility nationwide.
A total of 5,580 individuals filled out the Sonar-Brazil questionnaire. A total of
60 responses were excluded from the analysis: 39 did not meet the age criteria, 12
had missing or inconsistent data on weight and/or height (such as responses of ‘zero’
or ‘I don't know’), and 9 provided inconsistent data regarding daily routines. Therefore,
the final sample for Sonar-Brazil comprised 5,520 Brazilian adults. Night and shift
workers were also excluded from the present study due to their distinct circadian
rhythms and irregular sleep patterns, which could potentially confound the analysis
of sleep behaviors and their predictors (n = 260). Consequently, the current analysis was conducted with a sample of 5,260 Brazilian
adults.
To estimate population proportions with a 95% confidence level and a margin of error
of 5%, an initial minimum sample size of 385 valid questionnaires was determined.
However, the sample size was not fixed, and additional efforts were made to increase
it as much as possible, thereby reducing the margin of error. The final sample size
of 5,260 provides proportion estimates with a 95% confidence level and a margin of
error lower than 4%, enhancing the robustness of the findings.
Brazil is geographically divided into five macroregions—North, Northeast, Midwest,
Southeast, and South—each characterized by distinct climate, economic, and cultural
characteristics. According to the 2022 Brazilian Demographic Census, conducted by
Instituto Brasileiro de Geografia e Estatística (IBGE)[39], Brazil's total population is of 207,750,291 inhabitants, with regional distributions
as follows: North – 8.6%; Northeast – 26.7%; Southeast – 42.0%), South – 14.8%; and
Midwest – 7.9%). The final sample of the current reflects this demographic composition:
North – 9.1%; Northeast – 27.5%; Southeast – 41.2%; South – 14.3%; and Midwest – 7.9%).
Sex: Female and Male Subjects
In the present study, we opted to use the term sex rather than gender because sex is appropriate when focusing on physiological differences between male and female
individuals, such as hormonal profiles, reproductive anatomy, and chromosomal distinctions.
Since our aim was to analyze these biological variables, sex was the most precise term to describe the distinctions captured in our data, ensuring
clarity and scientific accuracy in the context of our research.
Data Collection
The participants filled out an online structured questionnaire after reviewing and
agreeing to sign the informed consent form. The questionnaire covered a range of sociodemographic
variables and lifestyle and health-related factors.
All data collection procedures were conducted in accordance with the principles outlined
in the Declaration of Helsinki and received approval from the institutional Ethics
in Research Committee.
Dependent Variable: BMI
The participants reported their weight and height in the questionnaire, and these
data were used to calculate the BMI, which was analyzed as a continuous and as a categorical
variable. We considered individuals with overweight or obesity (BMI ≥ 25 kg/m2) as having excess weight.[1]
Focal Independent Variables: Sleep Quality and Sleep Duration
We used the validated Brazilian Portuguese version of the Pittsburgh Sleep Quality
Index[16] (PSQI) to evaluate sleep quality. The questionnaire consists of 19 self-rated questions
organized into 7 components: subjective sleep quality, sleep latency, sleep duration,
habitual sleep efficiency, sleep disturbances, use of sleeping medication, and daytime
dysfunction. Each component is scored from 0 to 3, and the global score, derived from
the sum of these component scores, ranges from 0 to 21. Higher scores indicate poorer
sleep quality. The global score was categorized as good (≤ 5) or poor (> 5) sleep
quality.
Sleep duration (in hours) was calculated by subtracting usual wake time from bedtime,
regarding typical weekdays and weekends, with responses recorded in 30-minute intervals.
Weekly averages of waketime, bedtime and sleep duration were computed using the formula:
[(5 × weekdays value) + (2 × weekend value)] / 7.
Although sleep duration is one of the components of the PSQI, we opted to analyze
this variable separately as well, because it is an important and frequently studied
indicator of sleep quality. Based on the cutoff points established by the United States
National Sleep Foundation,[17] short sleep duration was defined as less than 7 hours per night.
Focal Mediating Variables
We hypothesized several lifestyle and health-related mediators, including: duration
of weekly physical exercise (measured in hours based on reported days and duration),
tobacco smoking (yes or no), alcohol intake (assessed by the total number of drinks
and frequency of consumption per week), screen time before bedtime (measured in minutes,
it includes the use of various devices, such as cell phones, televisions, computers,
and tablets), diagnosed depression and chronic conditions (including diabetes, hypertension,
hypercholesterolemia, heart disease, hypertriglyceridemia, and/or metabolic syndrome).
Another hypothesized mediator was diet quality, assessed using a food frequency questionnaire
that encompassed 17 categories. The participants reported their typical consumption
days from Monday to Sunday, specifying frequency and time. Each food group received
positive scores for healthy choices (such as, fresh fruits, vegetables, dairy) and
negative scores for unhealthy options (such as, sweets, fried snacks, sugary drinks).
The total score, ranging from 19 to 77, reflected diet quality and was categorized
into tertiles: low – 19 to 45; intermediate – 46 to 54/ and high – 55 to 77.
Lastly, we examined whether having dinner as the largest meal of the day acted as
a mediator. The participants identified their largest meal, defined as the one with
the highest caloric intake, with possible responses of breakfast, lunch, dinner, or
none.
Statistical Analysis
All analyses were performed using the Stata software, version 18 (StataCorp LLC).
The characteristics of the participants divided by sex were compared using the student's
t-test for continuous variables and the Chi-squared test for categorical variables.
We estimated the adjusted marginal probabilities of excess weight across sleep quality
groups, defined by PSQI scores < 5 (good), from 5 to 10 (poor), and > 10 (very poor),
for all participants, as well as separately for female and male subjects. Adjustments
were made for age, level of schooling, and marriage status. Additionally, restricted
cubic splines, adjusted for the same variables, were used to illustrate the association
between sleep duration and BMI for all participants, as well as for female and male
subjects separately.
To examine the association involving poor sleep quality, short sleep duration, and
both BMI and excess weight, we employed multiple linear and logistic regression models.
In the linear regression, BMI was treated as a continuous outcome, yielding β coefficients
and 95% CIs. Logistic regression was used to assess the odds of having excess weight,
providing odds ratios (ORs) and 95%CIs. Model 1 was adjusted for physiological factors
(age, depression, and chronic conditions); model 2 included adjustments for socioeconomic
and lifestyle factors (level of schooling, marriage status, smoking, alcohol intake,
and screen time before bed) in addition to the variables in model 1; and model 3 included
adjustments for dietary and physical activity factors (physical exercise, diet quality,
and dinner as the largest meal of the day). Statistical significance as p ≤ 0.05.
As a preparatory step for the mediation analysis, we first examined the unadjusted
regression coefficients to investigate the relationships regarding each hypothesized
mediator and the predictors (poor sleep quality and short sleep duration), as well
as the outcome (BMI). Given that we did not find associations involving short sleep
duration (β = 0.22; 95%CI: -0.24 01500.69; p = 0.348) or poor sleep quality (β = 0.37; 95%CI: -0.08–0.82; p = 0.109) and BMI among male subjects, the analysis of mediating factors was conducted
exclusively for the group of female participants.
We then conducted univariate analyses among female participants to test the associations
of each mediator with the outcomes and predictors, as shown in [Supplementary Table S1] (online only). As a result of this analysis, the mediators tested for the relationship
between poor sleep quality and BMI (all with p < 0.05 for both predictor and outcome) included: depression (percentage), chronic
conditions (percentage), smoking (percentage), alcohol intake (times per week), screen
time before bedtime (minutes), physical exercise (hours per week), diet quality (score),
and dinner as the largest meal of the day (percentage). For short sleep duration,
the mediators examined were depression (percentage), chronic conditions (percentage),
screen time before bedtime (minutes), and dinner as the largest meal of the day (percentage).
We used the Stata “medeff” package for the mediation analysis, calculating estimates
for the average causal mediation effect (ACME), which represents the portion of the
effect of the independent variable on the dependent variable mediated by the hypothesized
mediator. We also calculated the average direct effect, which reflects the independent
variable's effect on the dependent variable without mediation, and the percentage
of the effect mediated, indicating the proportion of the total effect explained by
the mediator. These calculations were performed using the quasi-Bayesian Monte Carlo
method based on a normal approximation with 5 thousand simulations. Statistical significance
was determined based on the 95%CI, with significant values highlighted when the interval
did not include zero. Variables with a significant ACME were considered mediators.
Results
A total of 5,260 adults (aged 18 to 65 years; 70% of female subjects) participated
in the current study. Significant sex-based differences (p ≤ 0.01) are detailed in [Table 1]. In the comparison between male and female subjects, the female participants had
a higher prevalence of higher education (73% versus 67%) and depression (15% versus
10%) respectively, while the male subjects were more likely to be married or in stable
unions (46% versus 42% of female participants). The male subjects presented higher
mean BMI, a greater prevalence of excess weight (61% versus 44%), and higher rates
of tobacco smoking (8% versus 5%), alcohol consumption (2.51 versus 1.59 times per
week), and physical activity (0.68 additional hours per week). They also presented
a higher percentage in the lowest diet quality tertile (39% versus 32%), were more
likely to report dinner as their largest meal (8.52% versus 8.33%) and presented later
bedtimes (23:53 versus 23:25). In contrast, the female subjects presented poorer sleep
quality (PSQI > 5: 53.39% versus 45.74%) and longer sleep durations (7.54 versus 7.28 hours).
Table 1
Participant characteristics by sex
|
Variables
|
All
(n = 5,260)
|
Female subjects
(n = 3,699)
|
Male subjects
(n = 1,561)
|
p-value[a]
|
|
Mean age (years)
|
36.39 ± 12.37
|
36.32 ± 12.48
|
36.55 ± 12.40
|
0.54
|
|
Age groups (years): %
|
|
|
|
|
|
18–29
|
36.18
|
36.09
|
36.39
|
0.94
|
|
30–49 years
|
46.16
|
46.31
|
45.80
|
|
|
50–65
|
17.66
|
17.60
|
17.81
|
|
|
Level of schooling: %
|
|
|
|
|
|
Lower education
|
28.71
|
26.93
|
32.93
|
< 0.001
|
|
Higher education
|
71.29
|
73.07
|
67.07
|
|
|
Marriage status: %
|
|
|
|
|
|
Single/Divorced/Widowed
|
56.60
|
57.75
|
53.88
|
0.010
|
|
Married/Living with partner
|
43.40
|
42.25
|
46.12
|
|
|
Mean BMI (kg/m2)
|
25.66 ± 4.93
|
25.27 ± 4.97
|
26.60 ± 4.69
|
< 0.001
|
|
Excess weight (BMI > 24.9kg/m2), %
|
49.18
|
44.23
|
60.92
|
< 0.001
|
|
Depression: yes (%)
|
13.57
|
15.11
|
9.93
|
< 0.001
|
|
Chronic conditions: yes (%)
|
26.16
|
25.63
|
27.42
|
0.17
|
|
Tobacco smoking: yes (%)
|
5.82
|
4.84
|
8.14
|
< 0.001
|
|
Mean alcohol intake (times/week)
|
1.86 ± 2.83
|
1.59 ± 2.44
|
2.51 ± 3.51
|
< 0.001
|
|
Mean screen time before bed (minutes)
|
3.10 ± 1.01
|
3.10 ± 1.02
|
3.12 ± 0.99
|
0.38
|
|
Mean physical exercise (hours/week)
|
4.15 ± 3.78
|
3.95 ± 3.64
|
4.63 ± 4.05
|
< 0.001
|
|
Mean diet quality score
|
49.17 ± 9.13
|
49.64 ± 9.04
|
48.06 ± 9.24
|
< 0.001
|
|
Low (1st tertile): %
|
34.30
|
32.22
|
39.21
|
< 0.001
|
|
Intermediate (2nd tertile): %
|
35.68
|
35.98
|
34.98
|
|
|
Good (3rd tertile): %
|
130.02
|
31.79
|
25.82
|
|
|
Largest meal (%)
|
|
|
|
|
|
Breakfast
|
5.80
|
6.84
|
3.33
|
< 0.001
|
|
Lunch
|
80.13
|
78.43
|
84.18
|
|
|
Dinner
|
8.38
|
8.33
|
8.52
|
|
|
None/Other
|
5.68
|
6.41
|
3.97
|
|
|
Wake time (decimal time
b
)
|
6.79 ± 1.18
|
6.79 ± 1.17
|
6.81 ± 1.23
|
0.49
|
|
Bedtime (decimal time
b
)
|
23.33 ± 1.21
|
23.25 ± 1.18
|
23.53 ± 1.26
|
< 0.001
|
|
Sleep duration (hours
b
)
|
7.46 ± 1.11
|
7.54 ± 1.11
|
7.28 ± 1.11
|
< 0.001
|
|
Short (< 7): %
|
29.94
|
27.66
|
35.36
|
< 0.001
|
|
Normal (7–9): %
|
64.16
|
65.34
|
61.37
|
|
|
Long (> 9): %
|
5.89
|
7.00
|
3.27
|
|
|
Sleep Quality (PSQI score)
|
6.06 ± 3.07
|
6.24 ± 3.12
|
5.62 ± 2.91
|
< 0.001
|
|
Poor (> 5): %
|
51.12
|
53.39
|
45.74
|
< 0.001
|
Abbreviations: BMI, Body Mass Index; PSQI, Pittsburgh Sleep Quality Index.
Notes:
a The p-values are derived from the student's t-test (for continuous variables) and from the Chi-squared test (for categorical variables). bWeekly average: ([weekday value * 5] + [weekend value * 2])/7. Significant associations
(p-values < 0.05) are shown in bold.
[Fig. 1] illustrates that the adjusted marginal probability of having excess weight (BMI > 24.9 kg/m2) increases as sleep quality declines, among all participants (PSQI < 5: 46%; PSQI
from 5–10: 50%; and PSQI > 10: 54%) and among the female subjects (PSQI < 5: 39%;
PSQI from 5–10: 46%; and PSQI > 10: 53%) (all p < 0.001). However, among the male subjects, the differences were minimal and not
statistically significant (PSQI < 5: 61%; PSQI from 5–10: 61%; and PSQI > 10: 60%;
p > 0.05).
Fig. 1 Marginal probability of excess weight (Body Mass Index [BMI] > 24.9Kg/m2) by sleep quality groups. Models are adjusted for age, schooling, and marriage status
(al l= 5,260; female subjects = 3,699; male subjects= 1,561).
[Fig. 2] illustrates similar findings, depicting the dose–response relationships between
sleep duration and BMI using restricted cubic splines. Among all participants and
specifically the female subjects, lower BMI values were associated with sleep durations
of approximately 8 hours, while the highest BMI values corresponded to shorter sleep
durations. In contrast, the association among males wass weaker and not statistically
significant ([Fig. 2]).
Fig. 2 Sleep duration and BMI. The red lines plot the predicted BMI values with 95% CIs
(color fill). Models are adjusted for age, schooling and marriage status (all = 5,260;
female subjects = 3,699; male subjects = 1,561).
[Tables 2], [3] present the associations involving short sleep duration and poor sleep quality with
BMI and excess weight respectively. Significant associations were observed across
all multiple regression models for both the overall sample and the female participants,
whereas no significant relationships were identified for the male participants. Among
the female subjects, short sleep duration was associated with an increase in BMI (model
3: β = 0.62; [95%CI: 0.27–0.97]; p = 0.001) and 21% higher odds of excess weight (model 3: 95%CI = 1.04–1.41; p = 0.014). Similarly, poor sleep quality was significantly associated with BMI for
the overall sample (model 3: β = 0.35 [95%CI: 0.10–0.61]; p = 0.007) and specifically for the female subjects (model 3: β = 0.46 [95%CI: 0.15–0.77];
p = 0.004). Additionally, the female participants with poor sleep quality had a 21%
greater likelihood of being classified as having excess weight (model 3: 95% CI = 1.05–1.39;
p = 0.01).
Table 2
Multiple linear regression models: association involving short sleep duration, poor
sleep quality, and Body Mass Index by sex
|
All
(n = 5,260)
|
Female subjects
(n = 3,699)
|
Male subjects
(n = 1,561)
|
|
β (95%CI)
|
p-value
|
β (95%CI)
|
p-value
|
β (95%CI)
|
p-value
|
|
Short sleep duration
|
|
|
|
|
|
|
|
Model 1
|
0.62 (0.33–0.90)
|
< 0.001
|
0.66 (0.31–1.02)
|
< 0.001
|
0.22 (-0.24–0.69)
|
0.34
|
|
Model 2
|
0.64 (0.36–0.93)
|
< 0.001
|
0.72 (0.36–1.08)
|
< 0.001
|
0.26 (-0.21–0.73)
|
0.27
|
|
Model 3
|
0.56 (0.28–0.84)
|
< 0.001
|
0.62 (0.27–0.97)
|
0.001
|
0.22 (-0.25–0.69)
|
0.35
|
|
Poor sleep quality
|
|
|
|
|
|
|
|
Model 1
|
0.49 (0.23–0.75)
|
< 0.001
|
0.65 (0.34–0.96)
|
< 0.001
|
0 0.37 (-0.08–0.82)
|
0.10
|
|
Model 2
|
0.46 (0.20–0.72)
|
0.001
|
0.59 (0.27–0.91)
|
< 0.001
|
0.43 (-0.03–0.89)
|
0.06
|
|
Model 3
|
0.35 (0.10–0.61)
|
0.007
|
0.46 (0.15–0.77))
|
0.004
|
0.36 (-0.10–0.81)
|
0.12
|
Notes: Short sleep duration is defined as < 7 hours, with ≥ 7 hours as the reference. Poor
sleep quality is defined as a Pittsburgh Sleep Quality Index score > 5, with a score
≤ 5 as the reference. Model 1 was adjusted for age, depression, and chronic conditions;
model 2 was adjusted for schooling, marriage status, smoke, alcohol intake, and screen
time before bed, in addition to model 1; model 3 was adjusted for physical exercise,
diet quality, and dinner as the largest meal of the day, in addition to model 2. Significant
associations (p-values < 0.05) are shown in bold.
Table 3
Multiple logistic regression models: association involving short sleep duration, poor
sleep quality, and excess weight (BMI > 24.9 kg/m2) by sex
|
All
(n = 5,260)
|
Female subjects
(n = 3,699)
|
Male subjects
(n = 1,561)
|
|
Odds ratio (95%CI)
|
p-value*
|
Odds ratio (95%CI)
|
p-value*
|
Odds ratio (95%CI)
|
p-value*
|
|
Short sleep duration
|
|
|
|
|
|
|
|
Model 1
|
1.24 (1.10–1.40)
|
0.001
|
1.22 (1.06–1.42)
|
0.007
|
1.90 (0.87–1.36)
|
0.44
|
|
Model 2
|
1.25 (1.11–1.42)
|
< 0.001
|
1.25 (1.07–1.45)
|
0.004
|
1.12 (0.90–1.40)
|
0.32
|
|
Model 3
|
1.22 (1.08–1.38)
|
0.002
|
1.21 (1.04–1.41)
|
0.014
|
1.12 (0.89–1.40)
|
0.34
|
|
Poor sleep quality
|
|
|
|
|
|
|
|
Model 1
|
1.15 (1.02–1.29)
|
0.019
|
1.28 (1.17–1.47)
|
< 0.001
|
1.03 (0.83–1.28)
|
0.79
|
|
Model 2
|
1.14 (1.01–1.28)
|
0.032
|
1.26 (1.09–1.45)
|
0.001
|
1.05 (0.84–1.32)
|
0.64
|
|
Model 3
|
1.10 (0.98–1.24)
|
0.10
|
1.21 (1.05–1.39)
|
0.01
|
1.05 (0.84–1.32)
|
0.67
|
Notes: Excess weight is defined as Body Mass index > 24.9kg/m2. Short sleep duration is defined as < 7 hours, with ≥ 7 hours as the reference. Poor
sleep quality is defined as a Pittsburgh Sleep Quality Index score > 5, with a score
≤ 5 as the reference. Model 1 was adjusted for age, depression, and chronic conditions;
model 2 was adjusted for schooling, marriage status, smoke, alcohol intake, and screen
time before bed, in addition to model 1; model 3 was adjusted for physical exercise,
diet quality, and dinner as the largest meal of the day, in addition to model 2. Significant
associations (p-values < 0.05) are shown in bold.
Since we only found significant associations between sleep and BMI among the female
subjects, we conducted mediation analysis solely in the females group, and the results
are presented in [Fig. 3] and [Supplementary Tables S2] and [S3] [online only], detailing the direct effect, the ACME, the total effect, and the
percentage of the total effect mediated by each lifestyle factor.
Fig. 3 Mediated total effect of poor sleep quality and short sleep duration with BMI among
female subjects. Models are adjusted for age, schooling and marriage status (n = 3,699). Abbreviation: ACME, Average Causal Mediated Effect. Notes: Short sleep duration is defined as < 7 hours, with ≥ 7 hours as the reference. Poor
sleep Quality is defined as a PSQI score > 5, with a score ≤ 5 as the reference. **
Statistically significant, as indicated by confidence intervals that do not include
zero. All models adjusted for age, schooling, and marriage status.
The results indicate significant direct and total effects of poor sleep quality on
BMI (direct effect: 0.46; 95%CI: 0.15–0.78). Several lifestyle and health-related
factors significantly mediate this relationship among female participants. Chronic
conditions demonstrated the highest mediation effect, accounting for 19.54% (95%CI:
12.48–44.76) of the total effect, with an ACME of 0.11 (95% CI: 0.05–0.18). Following
this, depression mediated 15.96% (95% CI: 10.12–36.11) of the total effect, with an
ACME of 0.09 (95%CI: 0.01–0.16). Diet quality accounted for 14.34% (95%CI: 8.90–34.59),
with an ACME of 0.08 (95%CI: 0.04–0.12). Screen time before bed mediated 11.30% (95%CI:
7.00–28.06), with an ACME of 0.06 (95%CI: 0.02–0.10). Lastly, dinner as the largest
meal of the day showed a mediation effect of 2.97% (95%CI: 1.77–8.59), with an ACME
of 0.01 (95%CI: 0.00–0.03) ([Fig. 3] and [Supplementary Table S2] [online only]).
Similarly, the direct and total effects of short sleep duration on BMI were significant,
with a direct effect of 0.62 (95%CI: 0.27–0.96). However, only screen time before
bed showed a significant ACME, of 0.03 (95%CI: 0.01–0.06), accounting for 4.93% of
the proportion mediated (95%CI: 3.23–10.61) ([Fig. 3] and [Supplementary Table S3] [online only]).
Discussion
As far as we are aware, the present is the first study to investigate sex differences
in the relationships between sleep and body weight status in Latin America, making
it pioneering in exploring the mediating factors involved in this relationship.
Our findings indicate that, among female subjects, poor sleep quality and short sleep
duration are significantly associated with higher BMI and increased odds of presenting
excess weight. In contrast, no significant associations were found among male participants,
supporting our hypothesis that the effects of sleep on BMI differ according to sex.
The mediation analysis further highlighted that chronic conditions, depression, and
diet quality play a key role in this relationship, with chronic conditions accounting
for 20%, depression, for 16%, and diet quality, for 14%. For short sleep duration,
screen time before bed was identified as the sole significant mediator, explaining
5% of the effect.
These findings are particularly significant in the context of the rising rates of
excess weight in Brazil. It is anticipated[2] that, by 2030, 68.1% of adults will be affected, with women expected to exhibit
a higher prevalence. This scenario is exacerbated by the growing economic impact of
overweight and obesity, which is projected to rise from US$1.96 trillion in 2020 to
over US$4 trillion by 2035–an economic burden comparable to the impact of the coronavirus
disease 2019 (COVID-19) pandemic in 2020.[1] Given these alarming trends, we hope the present study underscores the urgency of
developing comprehensive public health strategies that not only address traditional
risk factors but also integrate sleep hygiene and mental health interventions, with
particular attention to women in the prevention and management of excess weight.
Our results are consistent with those of previous research, such as the one by St-Onge
et al.,[13] who concluded, in their analysis of the Coronary Artery Risk Development in Young
Adults (CARDIA) study, that the associations between self-reported sleep duration
and body composition measures may be stronger in female than in male subjects. Similarly,
sex differences in the relationship between sleep duration and BMI were demonstrated
by Ren et al.[14] among Chinese adults: although ordinary least squares regression revealed no significant
associations for either sex after controlling for confounders, quantile regression
identified a U-shaped relationship for female subjects, indicating the lowest BMI
at ∼ 9 hours of sleep; in contrast, male subjects exhibited an inverse U-shaped relationship,
with BMI peaking at around 7 hours of sleep.
Despite previous studies[7]
[12]
[13]
[14]
[15] suggesting sex differences in the relationship between sleep and body weight, contrasting
results were found in a recent meta-analysis of 6 longitudinal cohort studies[7] with at least 12 months of follow-up, which showed that short sleep duration significantly
increased the risk of obesity in both male (OR = 1.26) and female subjects (OR = 1.36),
with no significant difference between the sexes.
Physiological and sociocultural factors are commonly mentioned to explain the significant
disparities in sleep quality and duration between male and female subjects.[18]
[19]
[20] Female subjects experience unique life stages–such as puberty, pregnancy, and menopause–that
heighten their vulnerability to sleep disorders due to hormonal fluctuations,[21] particularly sex steroids,[11] which influence sleep physiology, including reduced rapid eye movement (REM) sleep
during periods of elevated progesterone levels.[22] In addition to the more rapid accumulation of sleep debt,[23] women–particularly those with children–experience greater sleep disruptions, often
due to caregiving responsibilities.[5] Furthermore, they exhibit heightened vulnerability to psychosocial stressors and
are more predisposed to insomnia and nightmares following traumatic events.[24]
[25]
The current study supports these findings, showing that female subjects not only experienced
significantly poorer sleep quality but also exhibited a higher prevalence of depression
compared with the male participants. Notably, depression emerged as one of the strongest
mediators in the relationship between sleep quality and BMI. Depression can significantly
impact eating behaviors and has a well-established bidirectional relationship with
obesity[26] and sleep disturbances.[27] This cyclical link is likely driven by neurophysiological mechanisms, including
neurotransmitter dysregulation—particularly serotonin, norepinephrine, and dopamine—and
alterations in the hypothalamic-pituitary-adrenal axis, which is closely tied to the
circadian system.[28]
[29]
The additional mediators identified in the current study, such as diet quality, screen
time, and chronic conditions, are also well-documented in the literature as being
intricately linked to sleep and body weight through various interconnected pathways.
A poor-quality diet disrupts sleep and, in turn, poor sleep increases appetite, alters
energy expenditure, and affects hormonal regulation, raising the risk of excess weight.[30]
[31]
[32] Taheri et al.[33] found that habitual 5-hour sleep was associated with 15.5% lower leptin and 14.9%
higher ghrelin levels compared with 8-hour sleep, regardless of the BMI. Chronic conditions
impair sleep, disrupt metabolism, and cause hormonal imbalances, inflammation, and
stress, all contributing to weight gain.[34] Lastly, excessive screen time before bed negatively impacts sleep quality by suppressing
melatonin production due to blue light exposure, which not only affects the sleep-wake
cycle but also leads to appetite dysregulation and weight gain.[35] Engaging with stimulating content on social media can further delay sleep onset
and reduce sleep duration.[36] Additionally, chronic blue light exposure over time may disrupt circadian rhythms
and metabolism, increasing the risk of obesity.[35]
Strengths and Limitations
Strengths and Limitations
A strength of the current study is that the participants provided responses based
on their routines throughout the previous month, ensuring a genuine representation
of their lifestyle habits. Additionally, the survey also distinguished between weekdays
(workdays) and weekends (free days), calculating weighted aggregate scores to more
accurately reflect weekly eating and sleep patterns.
However, the present study has several limitations. First, the reliance on self-reported
questionnaires introduces the potential for underreporting or misreporting. Despite
this limitation, BMI calculated from self-reported data has been validated as a reliable
measure across various sociodemographic groups.[37] Although the PSQI is a validated and widely-recognized tool in epidemiological sleep
research, its subjective nature, relying on self-reports, presents a limitation to
the current study, as it lacks the objective sleep assessments provided by methods
such as polysomnography or actigraphy. Additionally, the cross-sectional design of
the study limits our ability to establish causality between sleep patterns and body
weight, highlighting the need for longitudinal research to further explore these relationships.
Conclusion
The current study revealed that poor sleep quality and short sleep duration are significantly
associated with higher BMI and increased odds of excess weight among female subjects,
while no association was found among the male participants. Chronic conditions, depression,
and diet quality mediated the effect of sleep quality on BMI, whereas screen time
before bed emerged as the sole significant mediator of the effect of short sleep duration
on BMI.
These findings highlight the importance of addressing sleep quality in interventions
aimed at managing BMI, particularly in female subjects. Future research should explore
the mechanisms behind these sex differences and mediators, using longitudinal designs
to clarify causal pathways and develop effective interventions.
