Sleep Duration Related to its Effect on Blood Pressure, Blood Glucose and BMI- A Systematic Review and Meta-Analysis

Objective: Sleep quality and its relation with optimal health is still controversial. So, The aim was to examine and gain insight into sleep duration and its impact on blood pressure, body mass index and blood glucose in adults. Methods: No RCTs were found between 1997 to 2019. A systematic review was then performed using cohort and cross-sectional studies including quantitative and qualitative data. Results: A total of 186921 patients from 19 eligible studies were included. Significant findings were seen for patients who slept <8 hours in association with hypertension (OR 0.88; 95% CI 0.86-0.91) as well as, obesity (OR 0.78; 95% CI 0.73-0.84). No significant difference was noted regarding sleep duration and impaired blood glucose. Conclusions: There is possibly a link between shorter hours of sleep (<8 hours a night) and the development of hypertension and obesity but not impaired glucose tolerance. Further research must be conducted with consistently reproducible methods and results in order to definitively conclude causation


Introduction
Sleep and its contribution to optimal health and well being have been the focus of clinical research for many decades but still, there is a lack of consensus on sleep quality and its duration. There is, however, satisfactory evidence that highlights the relationship between poor sleep and its impact on humans identifying the significant risk for developing hypertension, increased blood glucose and obesity [1]. Sleep arrears affect carbohydrate metabolism and endocrine function negatively, similar to that seen in normal ageing but with an increase in severity [2]. Using the benchmark of 8 hours of daily sleep the objective of this review is to relate the results of less than 8 hours of sleep with an increase in hypertension, blood glucose, and body mass index [3]. We hope that the results would support the insight into the duration of sleep and its role in changing the morbidity and mortality associated with the global epidemic of hypertension, diabetes mellitus, and obesity.

Summary of Methods
A systematic review of the existing English literature was performed followed by meta-analysis to study the association between lack of adequate sleep in adults and the risk of having higher blood pressure, blood sugar and BMI. This was done in concordance with the methods outlined by the Cochrane Handbook for Systematic Reviews of Interventions [4]. Volume 4; Issue 01 J Community Med Public Health, an open access journal ISSN: 2577ISSN: -2228 or "Diabetes mellitus" or "sleep loss" or "adult sleep disorders".

Selection of studies
All initially screened articles were entered in an excel sheet database. Six reviewers screened the title and abstract and full texts separately following inclusion and exclusion criteria. The studies that were included had to mention an effect or association of the duration of sleep in adults to two three variables mentioned above even if this was a secondary outcome in the study. The studies were randomly assigned to the reviewers and the studies were classified as "Include", "Exclude" or "Unsure". A third reviewer settled discrepancies.

Inclusion criteria
Study type: Randomized control trials, cohort and cross-sectional studies were included (no RCT was found). The studies were included if it mentioned sleep duration in hours and if it showed an effect or association with any of the three variables of interest (blood pressure, blood sugar, and BMI).
Population: Studies including adults (age 18 years or older). Some patients had sleep apnea. No further sleep disorder or psychiatric disorders were included.
Intervention and comparison related: Intervention and comparison were adults who slept less than 8 hours of sleep (this number was approximated depending on the study data) and adults who slept 8 hours or more.

Reported information (outcomes):
Studies that included data on three outcome variables: high blood pressure (or HTN), high blood sugar (or DM) and high BMI (overweight or obesity) were included. Obesity was defined as a body mass index (BMI) ≥30 kg/m 2 for men as well as for women [5]. Overweight was a BMI >25 kg/m 2 ) [5]. HTN was defined as a diagnosis of systolic blood pressure ≥140 mm Hg, or diastolic blood pressure ≥90 mm Hg [6]. Hyperglycaemia was defined as a diagnosis of DM (Diabetes mellitus), or capillary blood sugar ≥126 mg/dL if ≥8 h after the last meal; or capillary blood sugar ≥200 mg/dL if <8 h after the last meal [7]. Impaired glucose tolerance was defined as capillary blood sugar ≥140 mg/dL, 2 hours post glucose challenge.

Exclusion criteria
Non-original studies, structured abstracts, studies with unobtainable full text, case reports, case series, and commentary were excluded. Studies that included population less than 18 years old and people with psychiatric illnesses were also excluded. In addition, excluded studies were those that didn't show a correlation of sleep duration to the three variables of interest or studies without a specified number of hours. Studies that defined obesity by waist girth rather than BMI were also excluded.

Full-text screening
After reviewing the title and abstract by two reviewers independently, and a third reviewer to settle discrepancies if needed, the full text was reviewed further to ascertain eligibility and extract data.

Data extraction and management
Studies meeting the inclusion criteria were included for data extraction. A data extraction form was developed using Excel sheet and the following data were extracted from the full-text article: first author name, year of publication, journal name, hours of sleep reported, study population age and gender, number of people with the reported variables of interest (BMI, blood pressure and blood glucose), the total number of participants, the type of the study and the country in which the study was done.

Data Analysis (Quantitative and Narrative synthesis)
Each reviewer was assigned a specific number of journal articles to extract data. Both quantitative (means and odds ratios) and narrative data were extracted using two Excel sheets. The Excel sheets captured study findings in narrative form. The same variables were used: author's name, year of publication, journal name, hours of sleep reported, study population age, gender and the number of people with the reported variables of interest (BMI, blood pressure and blood glucose), the total number of participants and the type of the study conducted.
Review Manager Software (RevMan 5.3) [8] was used to conduct the analyses. For this review, the effect estimates were calculated odds ratio (OR) for dichotomous outcomes. Findings from studies that reported no data or OR were reported narratively. The I 2 statistic was reported as a measure of inconsistency. Heterogeneity was assessed by visual inspection of a forest plot along with consideration of the chi-squared test and the I 2 statistic. We interpreted the I 2 estimate as "might not be important" (0% to 40%), "moderate" (30% to 60%), "substantial" (50% to 90%) or "considerable" (75% to 100%) as recommended in the Cochrane Handbook for Systematic Reviews of Interventions [4].

Assessment of methodological quality of included studies
For observational studies (non-RCTs), the ROBINS-I (Risk Of Bias In Non-randomized Studies -of Interventions) criteria were used to assess the risk of bias in six fields: 1) Confounding factors, 2) Sampling, 3) Classification of intervention, 4) Bias due to missing data, 5) Bias in the measurement of outcomes, and 6) Bias in the selection of the reported results [9] as seen in Figures  1 and 2

Results
The search yielded 84 results and 65 were excluded after title/abstract and full text screening ( Figure 3). 19 studies were eligible for data abstraction. No RCT, 5 cohort studies and 13 cross sectional surveys were included. Quantitative data were extracted from studies (n=6) that outlined the number of participants who developed hypertension, diabetes/ glucose intolerance and changes in body mass index as a percentage of the total study population. Table 1 Shows the relation to sleep duration, three studies examined cases of hypertension (HTN), two studies gave data on diabetes/glucose intolerance and three studies submitted numbers for participants who fell into the obese category (BMI >30 kg/m 2 ). The total number of participants were 115720, aged between 18 to 83 years old. Included studies were from the USA, China, Korea and Saudi Arabia. Figures 4-6 show the forest plots that represent data comparing <8 hours sleep to ≥8 hours sleep. Obesity was a (BMI) ≥30 kg/m 2 for men as well as for women. Overweight was a BMI >25 kg/m 2 . HTN (hypertension) was defined as a diagnosis of systolic blood pressure ≥140 mm Hg, or diastolic blood pressure ≥90 mm Hg. Hyperglycaemia was defined as a capillary blood sugar ≥126 mg/ dL if ≥8 h after the last meal, or capillary blood sugar ≥140 mg/dL if <8 h after the last meal. Impaired glucose tolerance was defined as capillary blood sugar ≥140 mg/dL, 2 hours post glucose challenge. % expressed as a fraction of total population.   The rest of the studies (n=16) was recorded in a narrative summary table (Table 2). Seven studies had hypertension as an outcome, five reported BMI, two on diabetes, one examined both HTN and BMI and one on all 3 outcomes. The total number of participants for these studies was 71201 but it must be taken into consideration that information from two of those articles [11,12]

Discussion
Sleep quality is invariably tied into multiple aspects of one's health. From memory retention to mood stabilization, insulin resistance to blood pressure regulation, a good night's sleep is recommended to maintain homeostasis within the human body. But what exactly is a "good" night's sleep? It is not clearly defined by any known guideline. There are many theories and expert opinions about the topic, but the fact is that with age, our body's requirements for sleep vary. The National Institute of Neurological Disorders and Stroke suggest that most adults need 7 to 9 hours at night until the age of 60 where sleep duration may be reduced [30]. This systematic review aims to take a few aspects of health (blood pressure, blood glucose and body mass index) and examine their relationship with sleep duration in adults over 18 years old.
Due to the nature of the subject being studied, much of the evidence is taken from cohort and cross-sectional studies.
Since sleep deprivation is detrimental to the human body, it would be unethical to subject a participant to poor quality sleep intentionally. Most of the data in this systematic review, therefore, stemmed from self-reported surveys. Fang, et al., Najafian, Nouri & Mohammadifard and Li, et al. were among the largest crosssectional reports that were included, with over 10000 participants in each study [14,17,18]. All 3 papers investigated sleep duration and hypertension. We found that while most studies in our search collected data on hypertension, Body Mass Index (BMI) and diabetes, the figures were just a part of the population characteristics table and not necessarily the primary aim of the study to examine the relationship to sleep duration for all 3 outcomes.
The intrinsic flaw is that all the studies mentioned were not designed in the same way. Nor did they uniformly report their findings. This makes it difficult to compare and contrast their discoveries. Therefore, we constructed two tables to help display in an organized fashion the varying results. The quantitative table (Table 1) summarizes the results of 6 selected studies. Although the systematic review was determined to find the association of sleep duration with all 3 entities of obesity, hypertension and diabetes mellitus there was considerable heterogeneity across the selected studies regarding age of the population, data analysis, comparators and thus study conclusions. This is shown in the forest plots where I 2 was greater than 50%. One out of six studies delivered all three associations with sleep hours [11]. Five studies reported just one association [hypertension : Fang,et al. [14] and Guo,et al. [13]; obesity/overweight: Gangwisch,et al. [16], and Hwang,et al. [12]; diabetes mellitus: Gottlieb,et al. [15].
The commonest association of less than 8-hour sleep duration was found to be with obesity (BMI >30 kg/m 2 ) occurring in 3 out of 6 studies with the highest percentage of 17.71% found in Hwang et al total population of 6365 [12]. Conversely Brocato,et al. [11] showed that more than 8 hours of sleep had a higher percentage of obese patients (26.99%) out of 2686 participants in comparison to 8.89% in the less than 8 hours sleep category. Figure  5 shows obese cases related to sleep duration. The combined result is statistically significant favouring <8 hours sleep increasing the odds of a BMI >30 kg/m 2 .
The strongest association of hypertension with less than 8 hours of sleep was found in 3 out of 6 studies with the highest percentage of 29.14% in 9086 participants in Guo,et al. [13] Similar to the trend with BMI, in the category of more than eight hours sleep 23.05% out of 2646 population in Brocato,et al. [11] developed hypertension. In the largest study with 71,455 participants in Fang,et al. [14] the percentage of hypertensive cases associated with the greater than 8 hours of sleep group were comparable at 15.11% and 11.52% respectively. Due to the proportion of new events, in a weighted forest plot (Figure 4), there seemed to be a statistical significance that favours a longer sleep duration (>8 hours) as a possible protective factor for hypertension.
Finally, diabetic/hyperglycaemic association with less than 8 hours of sleep was found in 2 out of 6 studies. Interestingly, both Gottlieb,et al. [15] and Brocato,et al. [11] had a higher percentage of their population developing diabetes with greater than 8 hours of sleep (33.92% and 18.43% respectively) compared to the group that slept less than 8 hours. The forest plot in Figure 6 demonstrates though, a lack of statistical significance for this finding.
In Table 2 we extracted the odds ratios as a way of reporting the most relevant findings. 7 studies addressed sleep and hypertension. The general trend is that sleep that lasts less than 5 to 7 hours a day leads to an increased odds of developing hypertension. Li,et al. [18] showed an Odds Ratio (OR) of 1.24 (95% CI: 1.05-1.46) only for patients aged 18 to 44 years old with less than 7 hours of sleep a day. Najafian,Nouri & Mohammadifard [17] quoted an OR of 2.52 (95% CI: 2.17-2.93) for less than 5 hours of sleep a day. This finding was especially seen in women less than 60 years old [(OR=1.67, 95% CI: 1.29-2.16) for less than 5 hours sleep; (OR=1.28, 95% CI: 1.02-1.59) for 6-8 hours sleep]. Song,et al. [19] concluded that there was an increase in the odds of hypertension with less than 6 hours of sleep a day (OR=1.79; 95% CI: 1.05-3.03) after controlling for confounding factors (education, income, smoking, alcohol intake and physical activity). This study, however, targeted only pre-menopausal women in Korea who were not considered obese.
Notably, Gottlieb,et al. [21] found an increased odds for hypertension with less than 7 hours of sleep (OR=1.19, 95% CI: 1.02-1.39) as well as, 8-9 hours a night (OR=1.19, 95% CI: 1.04-1.37) and greater than 9 hours a night (OR=1.30, 95% CI: 1.04-1.62). These participants were initially part of a cohort study on cardiovascular outcomes of obstructive sleep apnea/hypopnea. This underlying comorbidity predisposes a person to disrupted or poor quality sleep so the authors adjusted for the apnea/hypopnea index as well as gender, ethnicity, age and BMI. This adjustment resulted in a decreased odds ratio across the board for all sleep duration categories in this study (<6, 6-7, 8-9 Aziz,et al. [23] investigated the sleep duration of participants in relation to morbid obesity (BMI >35 kg/m 2 ). 55% of the subjects were Hispanic. The authors found that for <6 hours of sleep the odds of morbid obesity were 1.8 (95% CI 1.5-2.2) but this result might have been skewed due to the ethnicity of the population that was investigated. Similarly, Hart,et al. [24] and Lauderdale,et al. [27] linked <6 hours sleep with an elevated BMI. In the former study [24], which was done as an online survey, 74% of the subjects were non-Hispanic. The disadvantage of the survey was that 13% of subjects had missing data. In the latter study [27] participants were taken from an existing ongoing cohort for the Coronary Artery Risk Development in Young Adults (CARDIA) study. These authors utilized a non-invasive wrist activity monitor in addition to the person's sleep log and predetermined activity score in order to denote sleep duration. This method however was dependent on the participant accurately starting and ending their sleep time.
Statistics from Iran have shown that the prevalence of obesity is on the rise. In 2015, it was estimated to be 21.7% for people above 18 years old [32]. Sepahvand,et al. [28]  They stated that there was a significant link to BMI but it was not well elaborated upon in their discussion.
Two cross sectional studies looked at the association of sleep with diabetes. The Taiwanese study Lin,et al. [29] detected a 5.24fold risk of diabetes for people 19-44 years old who slept <5 hours a day. In addition, Gottlieb,et al. [15] recorded an odds of 2.51 (95% CI: 1.57-4.02) for participants who slept ≤5 hours a night.
The comment must be made that all 3 outcomes (hypertension, glucose intolerance and BMI) are not only affected by sleep duration but in turn, they impact one another. Truncal obesity for instance, is tied to insulin resistance [33]. Diabetes accelerates vascular inflammatory processes and the resulting endothelial dysfunction manifests as hypertension [34]. Not to mention, ethnicity and nutrition further complicate the relationship. In isolation, it may be difficult to conclude that sleep deprivation leads to one outcome over the other. Though the sleep hours have definitive health value as evidently reported, the insight into these indicators is pending since sleep itself is a very complex and dynamic phenomenon [35,36].
We conclude that there is at least a possible link between shorter hours of sleep and the development of hypertension and obesity. There was no statistically significant association between sleep duration and impaired glucose tolerance thus far. If further research is done to explore the association of sleep duration and quality with hypertension, glucose intolerance and BMI, the study design must include an objective way to judge sleep duration, rather than self-reporting [37,38]. In addition, the results must somehow be adjusted to account for age, gender, ethnicity and diet.

Limitations
This systematic review has several limitations. Firstly, the studies used were ones that were published only in the English language. It is possible that some relevant literature was missed by not looking for articles that could have been translated. Secondly, the literature search included studies published from 1997-2019. Some full texts references data from the 1970s to 1980s but data prior to that were not included. Thirdly, some of the parameters were largely self-recorded, this relies on the participant accurately logging their sleep duration which could have varied based on the subject involved. The review was also done within a time constraint because of the availability of the team of authors to work on this project. In addition, the review team did not contain an experienced statistician or methodologist. The relative inexperience could have affected the accuracy of the interpretation of data.