Sex differences in the association between self-reported sleep duration, insomnia symptoms and cardiometabolic risk factors: cross-sectional findings from ELSA-Brasil CURRENT

Background: The U-shaped associations between sleep durations and cardiometabolic risk factors (glycated hemoglobin levels, obesity, hypertriglyceridemia, hypertension and HDL-Cholesterol levels) are still inconclusive. Moreover, as sleep is comprised of quantitative and qualitative aspects, exploring both insomnia symptoms and sleep duration are relevant when evaluating the potential effects of sleep problems on health. The aim was to evaluate sex-specific associations between sleep problems and cardiometabolic risk factors. Methods: This cross-sectional study used data from wave two of the Brazilian Longitudinal Study of Adult Health (ELSA-Brasil), including 7,491 women and 6,232 men. Questionnaires were administered to provide information about socioeconomic conditions, lifestyle, and sleep characteristics. A 12-hour fasting blood sample was drawn to measure serum cholesterol, triglycerides, and glycated hemoglobin. Blood pressure, weight and height were also measured using standard equipment. Generalized additive models were used to evaluate the curve shape of the relationship between self-reported sleep duration and the outcomes. Logistic regression was performed to investigate the magnitude of the associations of self-reported sleep duration, insomnia symptoms, and short sleep plus insomnia symptoms with cardiometabolic risk factors. Results: For women, self-reported sleep duration and insomnia symptoms (either separately or linked to short sleep duration) were associated with obesity, hypertension and glycated hemoglobin after adjusting for the confounders. The magnitudes of the associations between self-reported short sleep duration plus

insomnia symptoms and the outcomes were slightly increased, considering sleep duration or insomnia symptoms separately. For men, both long sleep duration and insomnia symptoms were associated with hypertriglyceridemia after adjusted for the confounders.
Conclusion: These findings suggest possible sex-specific patterns, since obesity, hypertension and high glycated hemoglobin were associated with self-reported sleep duration and insomnia symptoms in women, but not in men, and reinforce the importance of considering quantitative and qualitative aspects of sleep for the prevention and management of the outcomes. BACKGROUND Studies into sleep have increased in recent decades, since it was recognized as a public health issue due to the possible negative effects of sleep problems on health.
Concerning insomnia symptoms (difficulty initiating sleep, difficulty maintaining sleep, early morning awakening), the studies related to cardiometabolic issues are less extensive 3 , compared to the ones related to sleep duration. Having difficulty falling or staying asleep ('all of the time' or 'most of the time') was associated with type 2 diabetes in young female registered nurses, 18 and with both hypertension and dyslipidemia among public sector employees. 19 Meta-analysis studies showed that the risk of type 2 diabetes 7 and hypertension 3 were, respectively, 84% and 20% higher for subjects who reported difficulties in maintaining their sleep. 3,7 In a meta-analysis including studies on insomnia and obesity published in the past 10 years, the authors stated that associations are weak, particularly for European samples. 19 These findings stress that there is no single pattern of results and reinforce the need for studies with other population groups.
As sleep is comprised of quantitative and qualitative aspects, exploring both insomnia symptoms and sleep duration are relevant when evaluating the potential effects of sleep problems on health. In this context, the interaction between short sleep duration and insomnia symptoms, which leads to increased health risks has also been discussed. [20][21] Cai et al 21 found a significant association between insomnia symptoms and obesity among participants with short sleep duration, but not among those who reported normal sleep duration (6-8 hours). Similarly, short sleep duration with insomnia symptoms was associated with hypertension. 23 In relation to diabetes, an association was observed in those with chronic insomnia plus short sleep duration, but not among those with insomnia symptoms plus short sleep duration. 24 The above-mentioned associations do not specifically comment on the possible different effects between men and women, despite the known gender and sex differences in relation to sleep problems. [25][26][27] Few studies have investigated sexspecific associations between sleep and negative outcomes related to noncommunicable chronic diseases. 5 Gender differences have been detected in studies of sleep problems, with significant associations in women (but not men) with regard to hypertension, 28 the prevalence of dyslipidemia, 29 and obesity. 30 Short sleep duration was associated with a greater BMI and waist circumference among women but not among men. 31 However, as pointed out by Theorell-Haglöw et al. 32 research on gender differences in sleep medicine appears to be limited. In a metaanalysis study on sleep and type 2 diabetes, Shan et al. 5 emphasize the relevance of data on sex, which were usually limited in the original studies. These data reinforce the need for testing the hypothesis of sex-differences in the association between sleep patterns and cardiometabolic factors.
The mechanisms involved in the relationships between sleep problems and metabolic disorders include reduced glucose tolerance and compromised insulin sensitivity. 1 Insulin and glucose alterations associated with sleep restriction may be partially mediated by neuroendocrine changes, including high evening cortisol levels and increased sympathetic nervous system activity and catecholamine production. 33 Nocturnal sleep deprivation can lead to abnormal cortisol release during the night, which may result in decreased sensitivity to insulin in the morning.
Elevations in cortisol and increased sympathetic nervous system activity are also related to impaired vasodilatation and hypertension. Short sleep duration also reduces heart rate variability, 33

Variables definition
A comprehensive set of questionnaires was administered to provide detailed information on socioeconomic conditions, habits, lifestyle, and health. The study design, sampling procedures, construction of the questionnaire and quality assurance and control measures were previously detailed. [35][36] 1.
Sleep variables Participants who answered almost always or always were classified as having a complaint. Participants who reported any of the three described complaints were assigned to the insomnia group.

Statistical analyses
Descriptive analyses were performed to characterize the participants by sex. Given the heterogeneity in creating the range of reference categories for sleep duration, firstly Generalized Additive Models with Bernoulli distribution were used to evaluate the curve shape for the relationship between self-reported sleep duration and the outcomes and to determine the cut-off for hours of sleep duration. The models were fitted considering the potential confounders. After determining the cut-offs for selfreported sleep duration (≤ 6 hours; > 6 hours -≤ 8 hours; > 8 hours), logistic regression analyses were performed to investigate the magnitude of the associations between self-reported sleep duration, insomnia symptoms and the cardiometabolic risk factors (binary outcomes -high glycated hemoglobin levels, obesity, hypertriglyceridemia, hypertension and low HDL levels). The analyses were adjusted for potential confounders. Model 1 tested a crude association. Then, the analyses were adjusted for age and education (model 2) and included physical activity and depression symptoms (model 3). The fourth model was adjusted for self-reported sleep duration or insomnia symptoms to investigate independent effects. The last model included menopausal status. Joint associations of insomnia symptoms and self-reported sleep duration with all outcomes were examined considering participants with adequate sleep duration (> 6 hours -≤ 8 hours) and absence of insomnia symptoms as the reference group.
All analyses were performed using the software R version 2.15.

results
Participants in the ELSA-Brasil included 7491 women and 6231 men in this study.
For women and men, the mean age was 55.6 years and 55 years. In relation to sleep patterns, 47.6% and 49.1% of women and men, respectively, reported short sleep duration; 4.5% and 3.2%, respectively, reported long sleep duration; and 27.8% and 19.3% of women and men, respectively, reported insomnia symptoms. For both women and men, a higher proportion of individuals with self-reported adequate sleep durations was observed among those with a high level of education, high level of physical activity, and obesity and those without depressive symptoms. The group with insomnia symptoms was more likely to include participants with a low level of education, with no physical activity and with depressive symptoms (Table 1).
[ TABLE 1] For women, a U-shaped relationship was suggested between self-reported sleep duration and obesity, hypertension and glycated hemoglobin after adjusting for the confounders; adequate self-reported sleep duration (> 6 hours -≤ 8 hours) was not associated with cardiometabolic risk factors. In contrast, no significant association between self-reported sleep duration and the outcomes was observed for men, except for hypertriglyceridemia, which seems to be related mainly to long sleep duration ( Figure 1).

[FIGURE 1]
For women, in age-adjusted analyses revealed that self-reported sleep duration and insomnia symptoms were significantly associated with obesity, hypertension, low HDL, and high glycated hemoglobin levels compared to the respective reference groups. After adjusting for all potential confounders, the associations remained statistically significant for obesity, hypertension, and high glycated hemoglobin level, and the latter showed the strongest association. Also in women, it should be noted that for obesity, hypertension and high glycated hemoglobin levels, the magnitudes of the associations with the combination of self-reported short sleep duration with insomnia symptoms were higher than those observed for the factors analyzed separately ( Table 2). As previously highlighted in Figure 1, these associations were not statistically significant for men, except for hypertriglyceridemia; both long sleep duration and insomnia symptoms were associated with hypertriglyceridemia after adjusted for the confounders (Table 3).
[ -Sex-specific patterns considering short sleep duration Some studies reported significant associations between sleep problems and cardiometabolic outcomes for women but not men. Similar to our findings, in the Whitehall cohort, short sleep duration was associated with a higher risk of hypertension only in women. 28 Also, Grandner et al. 40 evaluated the association between hypertension and sleep duration in a large survey population and observed that short sleep was more closely related to hypertension in women than men. A meta-analysis of the association between sleep duration and hypertension revealed that the results from cross-sectional studies suggested statistically significant associations between short sleep duration and hypertension in women but not in men. 8 In relation to body weight, Westerlund et al. 30 observed higher BMI values in women and men with short sleep duration compared with adequate sleep duration (6-8 hours). On the other hand, similar to our findings, short sleep duration was associated with a greater BMI among women but not among men. 31 A recent metaanalysis 6 showed an association between short sleep duration and the risk for obesity (lowest risk at 7-8 hours sleep/day), stressing that the associations seem to be more evident in women than men. However, conclusions were not yet possible given the few studies addressing possible gender differences. 6 The same limitation was pointed out by Shan et al., 5  Concerning the U-shaped relationship between sleep duration and hypertension, a meta-analysis found that the results from prospective studies showed that neither short nor long sleep duration was statistically associated with the risk of incident hypertension. 8 However, similar to our findings for women, analyses from crosssectional studies showed that both short and long sleep durations were associated with prevalent hypertension in adults. 8 Another meta-analysis of cohort studies showed significant associations between short sleep duration and higher risk of hypertension. 3 However, the authors found no statistical evidence of any effect of long sleep duration on the risk of hypertension. 3 A meta-analysis of prospective studies identified that short sleep duration was significantly associated with increased risk of obesity, whereas long sleep duration had no effects on obesity. 16 A systematic review with meta-analyses examining the relationship between short sleep duration and several health outcomes identified significant results for diabetes, hypertension, cardiovascular disease and obesity. 11 The results from this same research group showed that long sleep duration was significantly associated with diabetes and obesity, but not with increased risk of hypertension. 12 In summary, the literature is not entirely consistent with respect to the U-shaped associations between sleep duration and metabolic outcomes. Some results from meta-analyses including longitudinal studies have showed statistically significant associations of both short and long sleep duration with diabetes and obesity, 5,12 while other studies that evaluated hypertension and obesity did not identify any associations. 3,16 Part of these disagreements could be attributed to methodological issues, such as the instrument used to assess sleep duration and the outcomes (subjective or objective measure), the cut-offs for sleep duration, and the duration of follow-up. 3 Sex-specific patterns considering insomnia symptoms Sex-specific patterns concerning the association between insomnia symptoms and cardiometabolic factors are scarce. Cross-sectional analyses 30 showed higher BMI values in women with insomnia symptoms, and a similar pattern but less marked among men. In contrast to our findings, insomnia was associated with a hypertriglyceridemia in women, but significant results were not found in men. 29 Also, difficulty maintaining sleep was associated with an increased overall risk for cardiovascular events in women, but not men 42 .
Regarding data from the general population without discussion of sex-specificities, insomnia symptoms were associated with hypertension medication, 44 dyslipidemia medication, 44 and type 2 diabetes. 7 A meta-analysis of cohort studies showed that insomnia symptoms were significantly associated with a higher risk of hypertension. 3 However, among individuals with adequate sleep duration, insomnia symptoms were not associated with obesity. 22 -Joint associations of insomnia and short sleep duration A recent study found a significant association between insomnia symptoms and obesity among participants with short sleep duration, but not among those who reported adequate sleep duration. 22 In the same way, short sleep duration with insomnia symptoms was associated with hypertension 23 and diabetes. 24 In a crosssectional study, associations between sleep duration and BMI were not modified by insomnia symptoms. 30 Our findings on the associations between self-reported sleep duration and each outcome independent of insomnia symptoms (and vice versa) also deserve attention. It has been discussed that insomnia and short sleep duration are distinct aspects of sleep that exhibit intersections. We did not find distinct results after adjusting for insomnia or sleep duration, i.e., the magnitude of the association was similar regardless of the adjustment for the other sleep variable. However, the combination of insomnia symptoms and self-reported short sleep duration appears to intensify the adverse effects of sleep problems separately from the independent associations. 20-21 Therefore, in line with previous studies, 23-24 we found that the magnitudes of the associations between self-reported short sleep duration plus insomnia symptoms and the outcomes were slightly increased. These findings stress the relevance of considering sleep problems as an additional risk of disease.

Possible explanations and mechanisms
In line with our findings, Prather et al. 27 showed that associations between sleep quality and biomarkers were stronger in women than in men. Poorer sleep quality was associated with 5-years increases in Interleukin 6, C-reactive protein, and fibrinogen in women but not in men. Therefore, causal pathways may be different for men and women. Therefore, causal pathways may be different for men and women. A laboratory study examining the effects of sleep deprivation on neural cardiovascular control in men and women found that sleep deprivation decreased muscle sympathetic nerve activity only in men, which in turn exerted a protective function on blood pressure. 43 Concerning differences in levels of reproductive hormones, it is possible that testosterone, which is higher in men, may attenuate the effects of poor sleep on inflammation. Supporting this hypothesis, lower circulating levels of estradiol contribute to the increased levels of inflammatory activity associated with poor sleep in women around the menopausal period (~55 years). 27 The association between sleep problems and greater psychological distress in women but not in men 26 could also explain these differences. Greater psychological distress in women is generally attributed to a more stressful role for women in Western societies.
Although sex-specific association between sleep and cardiometabolic problems represents an important research area, 5-6,32 the mechanisms underlying the different patterns are not completely elucidated because few studies have investigated sex differences related to this issue. 41 Therefore, further studies on sex differences and more mechanism studies are needed to clarify this issue.
In a general way, the biological evidence supporting a putative mechanism between sleep deprivation and cardiometabolic outcomes is derived from disturbances in autonomic function, and inflammatory and hormonal profiles. 33 For instance, sleep reduction affects energy balance, increases energy intake and reduces energy expenditure. Nocturnal awakening is also associated with altered leptin levels that lead to leptin resistance and result in glucose impairment. Elevated appetite could lead to weight gain and in turn, increases insulin resistance. 5,32 Possible pathways for the association between long sleep duration and cardiometabolic disease are not clear. Some authors discuss the possibility that long sleepers are previously sick individuals, and for that reason, these associations should be observed in this group. Long sleep is also associated with others sleep disorders, such as obstructive sleep apnea, and poor-quality sleep (increased sleep fragmentation and more frequent awakenings), which lead to changes in inflammatory markers that have been shown to be associated with metabolic dysregulation. 17,33,45 Disagreeing with the hypothesis that morbidity leads to long sleep duration, some studies stress the significant associations of long sleep duration with mortality, diabetes and other diseases, after controlling for other comorbidities. 17,[45][46] It should be noted that in the study of the joint associations of insomnia and sleep duration with diabetes, the adjustment for body mass index did not attenuate associations of long sleep duration without insomnia and diabetes. 17 In this way, a sensitivity analysis by exclusion of the participants diagnosed with myocardial infarction, stroke, and cancer during the first 2 years of follow-up did not change the association between long sleep duration and diabetes. 45 Clearly, there is a demand for further studies on long sleepers.

--Strengths and weakness
The strengths of this study are the large multicenter sample of middle-aged Brazilians, the use of standard equipment and techniques and the rigorous quality control of interviews and all measurements. All outcomes included were evaluated by objective measures. Also, we investigated the relationship between sleep duration and the cardiometabolic risk factors, examining firstly if the shape of these possible associations was linear or nonlinear in order to determine the cut off points for sleep duration based on the patterns of our sample.
The limitations of this study include the fact that all sleep variables were selfreported. Also, the insomnia symptoms did not include information related to daytime dysfunction. Although measuring objective sleep duration is more accurate, this procedure is generally not feasible in large epidemiological samples. 5,10 The current study demonstrated substantial reliability for insomnia symptoms and selfreported sleep duration questions. Also, objective sleep-duration measures seemed to be more strongly associated with hypertension than self-reported sleep duration, with no statistically significant differences. 3 Second, we did not have enough information to distinguish participants who were naturally short sleepers (short sleep duration but feel rested) from the others, and there may be differences between those groups. However, the investigation of individuals who reported short sleep duration plus insomnia symptoms showed results in the same direction of those observed for short sleepers. Therefore, we supposed that most individuals who reported short sleep duration do not have high sleep quality. Third, although the cross-sectional study design did not allow us to infer causality, the literature on sleep and metabolic disorders suggests that sleep problems most likely enhance the probability of developing cardiometabolic outcomes. 24 Fourth, our plasma measures were obtained based on standard procedures between 7:00 a.m. and 10:00 a.m., but the exact moment of assessment was not recorded. As we expect a diurnal variation for some biological markers, the timing of the blood test could be a potential confounder that was not considered. Fifth, we recognize the relevance of age in analysis of cardiometabolic factors. Nevertheless, our sample size (only ~4% reported long sleep duration) did not allow us to deeply study age effects -for

Consent for publication
Not applicable

Availability of data and materials
The datasets used and analysed during the current study are available upon request to the Study's Steering Committee, through an appointed representative, Dr.
Rosane Harter Griep (rohgriep@gmail.com). The ELSA-Brasil study, while open to any researcher, has a policy of requiring that all proposals of investigations pass through the study ś publications committee.

Competing interests
The authors declare that they have no competing interests    Figure 1 Generalized Additive Models (adjusted for all potential confounders) were used to evaluate th