A Systematic Review on Sleep Duration and Dyslipidemia in Adolescents: Understanding Inconsistencies

Although many questions about the role of sleep remain unanswered, it is known that sleep is not only a physiological function, but also performs an important role in promoting growth, maturation and general health of children and adolescents1, contributing significantly to cognitive, emotional functions and school performance2. Currently, there is a tendency for the young population to have irregular sleeping hours, with differences in bed and wake-up times between weekdays and weekends, especially as they get older2-4. 
 
There is a growing interest about the impact of sleep and its disorders on regulation of inflammatory processes and morbidities, particularly in the context of metabolic and cardiovascular diseases (CVD) and their complications1. In children and adolescents, cross-sectional5-7 and prospective8,9 studies have shown an association between overweight or obesity and few hours of sleep. In adults, there is evidence supporting this association, as well as correlations with insulin resistance, diabetes and cardiovascular diseases10-15. 
 
Few hours of sleep can also play a role in the etiology of a key risk factor to CVD, dyslipidemia12,14,15. Physiologically, sleep reduction is associated with hormonal alterations that may promote the development of an atherogenic lipid profile, including increase of cortisol and ghrelin and reduction of leptin levels, in addition to sympathovagal responses16-18. In order to obtain more information about the association between lipid metabolism alterations and sleep duration specifically in adolescents, we have performed a systematic review of the literature.


Introduction
Although many questions about the role of sleep remain unanswered, it is known that sleep is not only a physiological function, but also performs an important role in promoting growth, maturation and general health of children and adolescents 1 , contributing significantly to cognitive, emotional functions and school performance 2 . Currently, there is a tendency for the young population to have irregular sleeping hours, with differences in bed and wake-up times between weekdays and weekends, especially as they get older [2][3][4] .
There is a growing interest about the impact of sleep and its disorders on regulation of inflammatory processes and morbidities, particularly in the context of metabolic and cardiovascular diseases (CVD) and their complications 1 . In children and adolescents, cross-sectional 5-7 and prospective 8,9 studies have shown an association between overweight or obesity and few hours of sleep. In adults, there is evidence supporting this association, as well as correlations with insulin resistance, diabetes and cardiovascular diseases [10][11][12][13][14][15] .
Few hours of sleep can also play a role in the etiology of a key risk factor to CVD, dyslipidemia 12,14,15 . Physiologically, sleep reduction is associated with hormonal alterations that may promote the development of an atherogenic lipid profile, including increase of cortisol and ghrelin and reduction of leptin levels, in addition to sympathovagal responses [16][17][18] . In order to obtain more information about the association between lipid metabolism alterations and sleep duration specifically in adolescents, we have performed a systematic review of the literature.

Methods
This systematic review was based on the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement 19 .
The search was performed in the electronic databases Medline via Pubmed 20 Lilacs 21 , Web of Science 22 , Scopus 23 and Adolec 24 .
Selection of the descriptors used in the review process was made through MeSH (Pubmed's Medical Subject Headings). The search was performed in English, using three concept blocks: the first with terms related to sleep (sleep); the second with terms related to adolescence (adoles*, teen*, student*, youth, young); and the third with terms related to lipids (lipid*, lipemia*, cholesterol, HDL, LDL, triglyceride*, lipoprotein*, hypercholesterolemia*, hypercholesteremia*, dyslipidemia*, dyslipoproteinemia*, hyperlipidemia*, hyperlipemia*, "high density lipoprotein cholesterol", "low density lipoprotein cholesterol"). The Boolean operator "OR" was used for the combination of the descriptors within each block and the Boolean operator "AND" was used to combine the blocks amongst themselves. The truncation of terms was applied when necessary. No search limits were used for date, language, study design or sample size. The search was carried out in August 2014, contemplating articles published up to that date. Table 1 shows the search strategy used in each database.
Criteria for article inclusion in the systematic review were as follows: (a) studies on adolescents older than 10 years old; (b) studies that evaluated the association between sleep duration in hours and any lipid marker; (c) original research article. Articles evaluating any kind of sleep-related disorder, review studies, and experimental studies with animals were excluded. It was decided not to include theses, dissertations, and monographs. We reviewed the bibliographic references of reviews, systematic reviews, and meta-analyses that were found in the databases.
The articles were selected by two epidemiologists (GAA and LAB), initially based on title reading and then on abstract reading. Of the selected abstracts, the full articles were reviewed. In case of disagreement between the two reviewers with regard to the inclusion criteria, the title, and the abstract or the full article was maintained to be further evaluated. In case of disagreement with regard to the inclusion criteria, a third person was consulted.
Data from included articles were extracted independently, in duplicate (GAA and LAB), using a standard form. After extraction, data were compared and discussed. We extracted information about authorship, publication date, study place, population study, type of study, methods of sleep duration measurement and lipid profile assessment, sleep duration in hours, lipid markers, measure of association used to evaluate the correlation between hours of sleep and lipid profile, and variables used for adjustment of regression models.
We used an adaptation of the Newcastle-Ottawa (NOS) Quality Assessment Scale for Case-Control and Cohort Studies 25 , from the Ottawa Hospital Research Institute, to assess the quality of the longitudinal study included in this review. We also used the same scale adapted by Flynn et al 26 to assess the quality of cross-sectional studies. Web of Science (Topic(sleep*) AND Topic(adoles* OR teen* OR student* OR youth OR young) AND Topic(lipid* OR lipemia* OR cholesterol OR hdl OR ldl OR vldl OR triglyceride* OR lipoprotein* OR hypercholesterolemia* OR hypercholesteremia* OR dyslipidemia* OR dyslipoproteinemia* OR hyperlipidemia* OR hyperlipemia* OR "high density lipoprotein cholesterol" OR "low density lipoprotein cholesterol")) Scopus (TITLE-ABS-KEY(sleep*) AND TITLE-ABS-KEY(adoles* OR teen* OR student* OR youth OR young) AND TITLE-ABS-KEY(lipid* OR lipemia* OR cholesterol OR HDL OR LDL OR VLDL OR triglyceride* OR lipoprotein* OR hypercholesterolemia* OR hypercholesteremia* OR dyslipidemia* OR dyslipoproteinemia* OR hyperlipidemia* OR hyperlipemia* OR "high density lipoprotein cholesterol" OR "low density lipoprotein cholesterol")) Due to the great amount of methodological heterogeneity observed between the assessed studies, a narrative approach to synthesize the results of studies included in the present systematic review was considered a better strategy.

Results
The flowchart showing the selection process is shown in Figure 1. By the end of the evaluation process, of the 859 articles chosen after the removal of duplicates, 25 were submitted to full evaluation. Seven articles met the inclusion criteria at the end of the process. All studies used questionnaires to obtain hours of sleep. The variable "sleep duration" was used as continuous in three studies 27,29,30 ; whereas the other studies used different categories to classify sleep duration.
The methodological quality assessment of the seven included studies is shown in Table 3. Only two cross-sectional studies 28,31 obtained four points out of six in the bias risk evaluation. The longitudinal study showed a moderate risk of bias 27 . Table 4 shows the main results of the associations found and the control variables each study used. Considering the seven studies included, only in three an association was found between hours of sleep and lipid profile 27,28,33 . Two studies found that shorter sleep duration was associated with a worse lipid profile (total cholesterol and LDL-cholesterol) 27,28 , and the results of the third one 33 showed that long sleep duration was associated with high triglyceride levels. The other four studies 29-32 did not find any association.

Discussion
The present systematic review showed lack of consistent evidence regarding the association between sleep duration and lipid profile in adolescents. Few studies were found and some had methodological limitations. There was great heterogeneity regarding the classification and type of analysis of sleep duration and lipid metabolism markers, which probably contributed to the inconsistency of the observed results.
Concerning heterogeneity between studies, this systematic review included studies that evaluated the outcome using different methods (self-reported 27 , capillary blood sample 29 , venous blood sample 28,[30][31][32][33] or with different interval duration between the measure of exposition and the outcome 32 .
Gangwisch et al 27 did not exclude adolescents with dyslipidemia at baseline, thus, the incidence of dyslipidemia in adolescents could not be ascertained. Moreover, as the outcome established was self-reported, and the diagnosis of dyslipidemia depends on access to medical care, a bias may have occurred if adolescents from different socioeconomic status have different sleep habits. All studies included in this systematic review obtained information about sleep duration based on questionnaires, a method frequently used in sleep research because of its easy application and low cost. However, the validity of the information obtained through questionnaires is of concern, particularly when the tools have not been submitted to a validation process. Adolescents may report only socially desirable sleeping and waking up hours 34 . Although all studies used questionnaires, sleep duration evaluation was also heterogeneous: one study asked the parents about the adolescent´s sleep duration 31 , one used pre-defined categories of bedtime and waking-up time 32 , while the others asked about sleep duration in an open question [27][28][29][30]33 .
Actigraphy -based on monitoring of activities -has been established as a valid and reliable method to evaluate sleep-wake patterns in children, adolescents and adults 35,36 . Objective methods for hours of sleep quantification in a population-based study are difficult to use, particularly in studies with relatively large samples. Kong et al 28 used actigraphy in only about 7% of their study sample (138 out of 2,053) and demonstrated a reasonable agreement between actigraphy and adolescents' self-reports (intra-class correlation coefficient = 0.72, CI 95%: 0.61-0.80).
In the studies included in this review, duration of sleep was measured in two different ways, as a continuous 27,29,30 or categorical variable 28,31-33 . The lack of consensus about the best cut-off point to define short sleep duration makes it difficult to compare different studies, which would become easier if sleep duration were used as a continuous variable.
The present systematic review included a longitudinal study with important limitations and the cross-sectional studies showed associations in different directions. It was not possible to evaluate publication bias, due to the small number of studies identified. In summary, it is still uncertain whether there is an association between hours of sleep and lipid profile in adolescents. Heterogeneity regarding the way sleep hours were classified and analyzed, as well as the use of different lipids analytes may have contributed for the inconsistency of findings. More studies should be conducted on this issue to clarify the nature of this association and the involved biological mechanisms. These future studies must be longitudinal, use sleep duration as a continuous variable and consider the role of potential confounders or effect modifiers. Care must be taken to avoid over-adjustment, including variables that can be intermediary in the association between sleep duration and dyslipidemia such as BMI and food consumption.
Because of its strong association with cardiovascular disease in adults, it is important to identify and modify factors that are associated with lipid profile 15 in adolescents. If short sleep duration is responsible for an unfavorable lipid profile, interventions that improve the quality and duration of sleep may contribute to decrease long-term cardiovascular risk.