Association between sleep insufficiency and dyslipidemia: a cross-sectional study among Greek adults in the primary care setting

Objective To investigate the potential association between sleep insufficiency and dyslipidemia (DL) in the primary care setting using self-reported questionnaires. Material and Methods 957 adults aged between 19 and 86 years old from the rural area of Thrace, Greece were enrolled in this cross-sectional study. Multistage stratifed cluster sampling was used and the subjects were classifed into three groups according to sleep duration [short (<6h), normal (6-8h), and long (>8h) sleep duration]. DL was defined by a positive response to the question “Have you ever been told by a doctor or health professional that your blood cholesterol or triglyceride levels were high?”, or if they were currently taking antilipidemic agents. Sleep quality, utilizing Epworth sleepiness scale, Athens insomnia scale, Pittsburgh sleep quality index and Berlin questionnaire, was also examined. Results DL prevalence was significantly associated with short sleep duration (aOR=2.18, p<0.001) and insomnia (aOR=1.43, p=0.050), while its relation with poor sleep quality (aOR=1.31, p=0.094) and risk for obstructive sleep apnea (aOR=1.32, p=0.097) were of marginal statistical significance. Concerning insomnia subtypes, DL was significantly associated with difficulties maintaining sleep (aOR=2.99, p<0.001) and early morning awakenings (aOR=1.38, p=0.050), but not difficulties initiating sleep (aOR=1.18, p=0.328). Conclusion This study reveals an association between sleep pathology and DL. Thus, early pharmacological and cognitive or behavioral interventions that improve sleep are deemed necessary in order to decrease DL burden.


INTRODUCTION
Cardiovascular disease (CVD) comprises a major public health issue, as it constitutes the leading cause of noncommunicable disease mortality worldwide 1 . CVD pathogenesis is complex including non-modifiable risk factors, such as age, gender, race and ethnicity, and modifiable risk factors, such as smoking, lack of physical activity, diabetes mellitus, hypertension, and visceral obesity 2 . Dyslipidemia (DL) defined as high levels of total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), triglycerides (TG), and low levels of high-density lipoprotein cholesterol (HDL-C), is another wellestablished reversible risk factor for CVD, thus encompassing an important issue in the field of health promotion and CVD prevention 3, 4 .
Serum lipid and lipoprotein levels are influenced by lifestyle parameters, such as smoking, alcohol consumption, and exercise 5,6 . Due to public health concerns about the declining quality and quantity of sleep in modern society, research into the effect of sleep on health has blossomed. Although in recent years, both short and long sleep duration have been associated with chronic diseases such as hypertension 7 , diabetes mellitus 8 , and CVD 9 , results on the potential relationship between sleep duration and DL have been contradictory so far, as U-shaped associations 10,11 or relationships between DL and only short 12 or long sleep duration 13 have been presented. Apart from that, few studies have investigated the effect of sleep quality on DL reaching antithetical conclusions, as well [14][15][16] . Furthermore, there exists no consensus on the probable gender-specific association between sleep pathology and DL 17,18 . Finally, few European studies have investigated the potential association between sleep pathology and DL 12,19 . Our research group utilizing self-reported questionnaires has recently exhibited that sleep pathology is associated with increased prevalence of anxiety 20 , depression 21 , diabetes mellitus 22 , and hypertension 23 . In this paper, we aimed to investigate possible correlations between sleep quantity/ quality and DL considering several socio-demographic characteristics, lifestyle habits and health related characteristics of the participants and focusing on potential gender-specific associations.

Study sample, research design and covariates
The study population in this cross-sectional study consisted of 957 participants, 439 (45.9%) males and 518 (54.1%) females, with a mean age of 49.62±14.79 years (range, 19-86 years; median age, 50 years). The research design of this study is reported in Serdari et al. (2020) 20 . The questionnaires used in order to collect standard socio-demographic characteristics, lifestyle, and dietary habits and health related characteristics of the participants are reported in Matziridis et al. (2020) 22 . Chronic disease morbidity was defined as the self-reported preexisting (for ≥3 months over the past year) health problems, such as: hypertension, diabetes mellitus, cancer, cardiovascular, rheumatic, gastrointestinal, pulmonary, or neurologic disease.

ETHICS
All procedures performed in the study were in accordance with the ethical standards of the Democritus University Ethics Committee, which approved its conduct, and with the standards of the Helsinki declaration (1964) and its later amendments. Informed consent was obtained from all participants of the study.

Estimation of sleep duration and sleep efficiency
Participants answered the following questions: "At what time do you normally go to bed?", "At what time do you normally get up?", and "On average, how many hours do you sleep every night?" for an average weekday and weekend day over the previous month. Time in bed, sleep duration, and sleep efficiency calculation formulae on weekdays, weekend days and weighted mean measures utilized in this study are reported in Matziridis et al. (2020) 22 . According to calculated sleep duration, participants were then classified into: short sleepers (<6 hours), normal sleepers (6-8 hours), and long sleepers (>8 hours) 24,25 .

Evaluation of sleep quality
Sleep quality was assessed with the Greek versions of Epworth sleepiness scale (ESS) 26 , Athens insomnia scale (AIS) 27 , Pittsburgh sleep quality index (PSQI) 28 , and Berlin Questionnaire (BQ) 29 that evaluate excessive daytime sleepiness, insomnia, sleep quality, and risk of obstructive sleep apnea (OSA), respectively. With regards to insomnia characteristics, participants were asked whether they experienced difficulties initiating or maintaining sleep or early morning awakenings.

Definition of DL
DL was defined by a positive response to the question "Have you ever been told by a doctor or health professional that your blood cholesterol or triglyceride levels were high?" or if they were currently taking antilipidemic agents 30,31 .

Statistical analysis
Statistical analysis of the data was performed using IBM Statistical Package for the Social Sciences (SPSS), version 19.0 (IBM Corp., Armonk, NY, U.S.A.). The normality of quantitative variables was tested with Kolmogorov-Smirnov test. Quantitative variables were expressed as mean±standard deviation (SD) and qualitative variables were expressed as absolute and relative (%) frequencies. In particular, mean estimated time of sleep characteristics (i.e., bedtime, rise time, time in bed, and sleep duration) were expressed as HH:MM. We conducted the following analyses: (i) in the univariate analysis, the association of DL with subjects' characteristics, sleep characteristics, and sleep disorders were assessed using the chi-square test and Student's t-test; (ii) multivariate stepwise logistic regression analysis was used to explore the independent risk factors for DL, controlling for all subjects' characteristics; (iii) for the evaluation of the effect of sleep duration and sleep disorders on the prevalence of DL, two different logistic regression models were constructed: model 1 (crude, unadjusted) and model 2 (adjusted for subjects' socio-demographic characteristics: gender, age, marital status, cultural status, place of residence, education level, working status, financial status; lifestyle habits: smoking status, alcohol consumption, daily coffee consumption, caffeine consumption in the evening, adherence to the Mediterranean diet, time watching TV or using a computer before bedtime, physical activity, nap during the day; and health related characteristics: subjective general health status, BMI, chronic disease morbidity, anxiety, depression, and use of sleep medication). Odds ratios (OR) with their 95% confidence intervals (CI) were estimated as the measure of the above associations.
Receiver operating characteristic (ROC) analysis was used to provide the ability of sleep duration to classify subjects with DL. The area under the ROC curve (AUC), sensitivity, and specificity were estimated. The optimal cutoff value of the sleep duration that differentiates DL from non-DL individuals was derived according to Youden index. All tests were two tailed and statistical significance was considered for p-values≤0.05.

Subjects' characteristics
Participants' socio-demographic, lifestyle and health related characteristics are summarized in Tables   average self-reported sleep duration was 6hrs and 19min on the weekdays and 6hrs and 45min on the weekends; 31.7% and 22.9% of the sample reported sleep duration less than 6hrs, while 7.9% and 14.2% of the sample reported sleep duration more than 8hrs per night on the weekdays and on the weekends, respectively. Sixty-six participants (6.9%) were on sleep related medications. A total of 311 participants (32.5%) were classified with DL. The prevalence of DL in relation to participants' socio-demographic, lifestyle and health related characteristics is summarized in Tables 1 and 2. Significant determinants of DL obtained by multivariate logistic regression models are presented in Table 3. duration there were more than 3-times higher odds for DL compared to subjects with normal sleep duration (OR=3.26, p<0.001). A 2.65-fold (p<0.001) and a 3.92-fold (p<0.001) increase in odds of DL was associated with short sleep duration in women and men, respectively.

Independent effect of DL on sleep habits
Two separate multivariate logistic regression models, controlling for the effect of all subjects' socio-demographic, lifestyle, and health related characteristics, were constructed in order to assess the independent effect of sleep duration on the prevalence of DL. When sleep duration was entered in the model as a continuous variable, it remained a statistically significant independent determinant of DL (p=0.049); in particular, shorter sleep duration by one hour was associated with a 14%-increase in the odds for DL (aOR=1.14, 95% CI=1.00-1.30).
When sleep duration was entered in the multivariate logistic regression model as a categorical variable, the inverse relationship between DL and sleep duration persisted even after the adjustment for all subjects' characteristics. In particular, the odds of DL were more than 2 times higher in subjects sleeping less than 6 hours (aOR=2.18, p<0.001) relative to those with normal sleep duration; the respective odds for DL were similar in the two genders (aOR=2.01, p=0.023 in females; aOR=2.84, p<0.001 in males). Sleeping longer than 8 hours showed no significant effect on the development of DL (Table 5).

DL and sleep disorders
According to the Greek versions of ESS, AIS, PSQI and BQ the prevalence of daytime sleepiness was 8.7% (83 subjects), insomnia 18.0% (172 subjects), poor sleep quality 38.5% (368 subjects), and high risk of obstructive sleep apnea 36.4% (348 subjects). The internal consistency of all four questionnaires was very high (Cronbach α coefficient was ranged from 0.74 to 0.88). The presence of DL in relation to sleep disorders is shown in Table 6.
Univariate statistical analysis showed that DL was more frequent in subjects with insomnia (40.7% vs. 30 Notes: aOR = Adjusted odds ratio; CI = Confidence interval; all subjects' sociodemographic, lifestyle habits and health related characteristics were included in the model; All variables (with the exception of age) were binary (no, yes); category "no" forms the reference group.

DL and sleep habits
The association of DL with subjects' sleep characteristics is shown in Table 4. The average weekly time in bed and sleep duration were calculated and compared between the two groups. It was noted that, although subjects with DL spent 11min longer time in bed (p=0.007), they reported a 22min shorter sleep duration (p<0.001) and lower sleep efficiency (p<0.001) compared to those without DL. All the above relations between DL and sleep characteristics remained unchanged among females and males. In particular, females with DL used to sleep 21min less than females without DL (p=0.005) and males with DL used to sleep 22min less than males without DL (p=0.003). Among subjects with DL, all three sleep characteristics were similar between males and females (time in bed: p=0.085; sleep duration: p=0.303; sleep efficiency: p=0.964).
In the sequence, according to the self-reported sleep duration, participants were categorized into three groups: short (<6h), normal (6-8h), and long (>8h) sleep duration. The association between the development of DL and sleep duration, which was considered as a categorical variable (   Table 5. Association of sleep duration with dyslipidemia (DL) using logistic regression models.
Notes: cOR = Crude odds ratio; aOR = Adjusted odds ratio; CI = Confidence interval; Model 1 = Crude, unadjusted; Model 2 = Adjusted for socio-demographic characteristics, lifestyle habits (smoking status, alcohol consumption, daily coffee consumption, caffeine consumption in the evening, adherence to the Mediterranean diet, time watching TV or using a computer before bedtime, physical activity, nap during the day) and health related characteristics (subjective general health status, BMI, chronic disease morbidity, anxiety, depression, and use of sleep medication).

DISCUSSION
A population based cross-sectional study utilizing selfreported questionnaires was conducted in the primary care setting in northeastern Greece and presented significant associations between short sleep duration and DL among adults irrespective of gender. Moreover, it was revealed that DL prevalence was significantly associated with insomnia, particularly with difficulties maintaining sleep and early morning awakenings.
The high-evidenced prevalence of DL (32.5% of the study population sampled) is consistent with the results of Merekoulias et al. (2014) 32 who reported a prevalence of 35.7% of DL among individuals over 70 years old with poor sleep quality in Western Greece. Furthermore, DL was more prominent among Greek Muslims affecting 39.2% of our population, compared to Expatriated Greeks and Greek Christians with proportions of 28.8% and 29.9%, respectively. Several studies have also underlined the growing prevalence of DL in minority groups 33,34 . Moreover, rural citizens presented DL twice more frequently when compared to inhabitants of urban areas. This finding probably derives from educational and socioeconomic disparities; hence, people with low educational status from urban areas seem to be more prone to the detrimental effect of DL. Indeed, low financial and educational status in a Korean population favored the development of DL 35,36 . Quite interestingly, the marital status affected the distribution of DL in our sample, with the widowed patients holding the lion's share, as also proved in the research of Nikparvar et al. (2020) 37 . Additionally, our results demonstrated that factors such as ex-smoking, obesity, sedentary life, and unhealthy diets may contribute to higher DL rates of developing populations, along with previous studies 38,39 . Finally, in keeping with our results emerging evidence from the study of Rekleiti et al. (2012) 40 demonstrated the independence of DL as a prognostic factor for anxiety.
Available cross-sectional studies have not reached a consensus concerning the relationship between sleep duration and DL. For example, Lin et al. (2017) 10  Our study did not elicit gender-specific associations between short or long sleep duration and DL. In keeping with our results, Song et al. (2020) 18 could also not find significant interactions between sleep duration and sex with respect to abnormal serum lipid levels. In contrast, data from the China Health and Nutrition Survey (2009) showed that both short and long sleep duration were associated with higher risks of abnormal serum lipid profiles in women, but not in men 17 . Similarly, Kaneita et al. (2008) 43 found that both short and long sleep duration were positively associated with high TG and low HDL-C in Japanese females, but not males. Nakanishi et al. (1999) 44 could also not find a significant association between sleep duration and serum lipid and lipoprotein levels among Japanese men. Finally, Williams et al. (2007) 45 exhibited that short sleep duration was related to low HDL-C levels in adult American women with type 2 diabetes. Several investigators have tried to explain the gender-specific effects forging the relationship between sleep quantity and DL by implicating bias in self-reporting of sleep 46 and differences born either by social and household statutes 47 or by sex hormones variation 48 .
With regards to sleep quality, DL remained significantly associated only with insomnia (aOR=1. 43   utilizing data from more than 10.000 Chinese individuals, apart from 25% increased odds of elevated TC level only among women experiencing insomnia, did not find significant associations between insomnia and LDL-C, HDL-C, or TG levels among both sexes. In contrast, Rouleau et al. (2017) 49 exhibited that upon completion of an exercise-based cardiac rehabilitation program, greater improvement in insomnia symptom severity was associated with greater improvements in TC levels. Finally,   50 revealed that Taiwanese insomniacs had, surprisingly, lower TC levels compared to non-insomniacs.
Several mechanisms have been implicated in the pathogenetic relationship between inadequate sleep and DL. The study of Corgosinho et al. (2013) 51 highlighted the burden of short sleep duration on human metabolism, thus inducing increased dietary fat consumption, DL and obesity, which is many times difficult to confront especially among adolescents. For example, sleep restriction strongly influences metabolic hormones that regulate energy balance, as it has been shown that short sleep duration lowers the blood concentration of leptin that suppresses appetite, and increases the blood concentration of ghrelin, which promotes appetite and sequentially daily dietary intake of cholesterol, trans-fats, and saturated fats 52 . Apart from that, inadequate sleep has been associated with excessive daytime fatigue and consequent reduced engagement in physical activities. Low levels of physical activity have been shown to increase LDL and lower HDL levels 53 . Moreover, it has been presented that short sleep duration by increasing acute stress significantly raises TC and LDL-C levels 54 . In agreement, the evidence of Andersen et al. (2009) 55 on animal models suggest that the exposure to chronic stressful stimuli resulted in impairment of all parameters of the lipid profile. Finally, it has been proposed that sleep restriction could modify genetic risk factors for adverse blood lipid profiles 56 .
Our analysis manifests several strengths, as it is based on data from a large representative sample of the population in northeastern Greece that provided excellent response rates to sleep quality and quantity and DL measurements (71% compared, for example, to 63% in the Hordaland Health Study) 12 . Moreover, our sampling scheme ensured that the sample was randomly selected and representative of the general population of this area. Limitations of this study lie in the recall bias of self-reported sleep duration as reported sleep duration may be overestimated 57 . Nevertheless, self-report assessments of sleep have been shown to be valid measures compared to quantitative ones based on actigraphy 58 . Apart from that, data were not obtained separately on hypercholesterolemia or hypertrygliceridemia or in patients who were on treatment for dyslipidemia, and therefore the relationship between sleep disturbances with these two dyslipidemic states or patients who were on lipid-lowering agents could not be evaluated. Another limitation involves the properties of the cross-sectional study as causal relationships between sleep pathology and DL could not be determined.
Heterogeneity in sleep duration definition combined with the limited number of relevant studies performed 59,60 constitute the evidence based statistical interpretation a tedious task 61 . Thus, further prospective studies are required and a longitudinal study on this topic by our research team is scheduled.

CONCLUSION
To the best of our knowledge, this is the first conducted cross-sectional study in Greece that elucidates an association between short sleep duration, insomnia and DL. Considering its strengths and limitations it could be proposed that early pharmacological, cognitive, and behavioral interventions that improve sleep quality and quantity might serve as preventative measures for DL.

ACKNOWLEDGMENTS
The contributions of all of the participants, patient advisers, and interviewers are gratefully acknowledged.