A Meta-Analysis of a Cohort Study on the Association between Sleep Duration and Type 2 Diabetes Mellitus

Objective To study the association between sleep duration and the incidence of type 2 diabetes mellitus (T2DM) and to provide a theoretical basis for the prevention of T2DM through a meta-analysis. Methods PubMed, Web of Science, Scopus, Embase, Cochrane Library, ProQuest, CNKI, Wanfang, VIP, and SINOMED were searched from their inception until May 2020. All cohort studies on the relationship between sleep duration and T2DM in adults were included. According to the inclusion and exclusion criteria, two authors independently assessed the literature and extracted the data. Metaregression and publication bias were evaluated, and sensitivity and meta-analyses were conducted with RevMan 5.3. Results A total of 17 studies were collected, involving 737002 adults. The incidence of T2DM was 4.73% in short sleep duration (SSD) (t ≤ 6 h), 4.39% in normal sleep duration (NSD) (6 h < t < 9 h), and 4.99% in long sleep duration (LSD) (t ≥ 9 h). The meta-analysis demonstrated that SSD increased the risk of T2DM compared with NSD (RR = 1.22, 95% CI: 1.15-1.29, P < 0.001), LSD increased the risk of T2DM compared with NSD (RR = 1.26, 95% CI: 1.15-1.39, P < 0.001), and the risk of T2DM has no significant difference between SSD and LSD (RR = 0.97, 95% CI: 0.89-1.05, P = 0.41). The sensitivity of each study was robust and the publication bias was weak. Conclusion SSD or LSD can increase the risk of T2DM.


Introduction
Diabetes mellitus (DM) is an epidemic disease in recent years. According to the International Diabetes Federation Diabetes Atlas Ninth Edition published in 2019, it is estimated that 463 million adults (aged between 20 and 79 years) have DM worldwide, and the prevalence has reached 9.3%. This number is expected to reach 578 million ( DM is one of the leading causes of retinopathy, vascular disease, neuropathy, amputation, heart disease, kidney failure, and premature death [3]. There are many risk factors for type 2 diabetes mellitus (T2DM), such as genetic factors and an unhealthy lifestyle. Studies have shown that sleep quality and sleep duration are also risk factors for T2DM [4][5][6]. However, relevant research conclusions are inconsistent. Some systematic reviews have indicated that short sleep or long sleep are risk factors for T2DM in adult [7][8][9][10] and risk factors for gestational diabetes mellitus in pregnant women [11][12][13]. However, studies for adults have not been updated in time. Therefore, we have conducted a metaanalysis of cohort studies to exhibit the relationship between sleep duration and T2DM in adults to provide the basis for primary prevention of T2DM.

Study Eligibility and Selection
Criteria. Two authors independently determined study eligibility. Any differences in opinion about eligibility were resolved through another author as a third-party consensus. The inclusion criteria were as follows: (1) cohort studies on the relationship between sleep duration and T2DM in adults, (2) reported sleep duration, (3) diagnosed T2DM, and (4) published English and non-English papers. Studies were not included if (1) they are cross-sectional studies, case-control studies, case reports, and review commentaries; (2) they do not report the incidence of T2DM; (3) they have participants < 10; (4) the subjects were special groups, such as pregnant women and patients after organ transplantation; (5) sleep duration was not reported clearly; (6) they were a duplicate report; and (7) they reported incomplete data and the relevant data were not available.  hours as a reference, it is categorized as short sleep duration (SSD) (t ≤ 6 hours), normal sleep duration (NSD) (6 hours < t < 9 hours), and long sleep duration (LSD) (t ≥ 9 hours). Asante et al. (2020) [15]. For all studies, we extracted the first author, publication year, study design, study location, study period, population, number of T2DM, age, gender (male/female), and data acquisition.

Assessment of Risk Bias.
The risk bias of the included studies was independently assessed by two authors. The cohort study was evaluated by the Newcastle-Ottawa Scale (NOS), which included eight items, categorized into three groups: the selection of study groups, comparability of groups, and ascertainment of outcome [16]. Each study will be evaluated by eight items, and high-quality choices were identified with a star. There are a maximum of one star for each high-quality item within the selection and outcome categories and a maximum of two stars for comparability.
2.6. Statistical Analysis. The meta-analyses were conducted using Review Manager software, version 5.3 (https:// community.cochrane.org/help/tools-and-software/revman-5). Dichotomous outcomes eligible in each study are reported as a risk ratio (RR) with an estimated 95% confidence interval (CI). Continuous outcomes are shown as the weighted mean difference (WMD) with the 95% CI, which were calculated from the mean, standard deviation (SD), P value, and sample size in each study. Heterogeneity was assessed using Higgins I 2 , which evaluates the percentage of total variation across studies that were due to heterogeneity rather than by chance. Thus, if I 2 > 50%, which was considered to reflect substantial heterogeneity, a random effect model was used. If I 2 ≤ 50%, which was considered to reflect no heterogeneity, a fixed effect model was employed. The chi-square tests were also used to evaluate the heterogeneity: P < 0:1 indicates heterogeneity, while P > 0:1 indicates no heterogeneity. Based on   Journal of Diabetes Research clinical knowledge, the study location and study period were considered to be responsible for heterogeneity, and so, these parameters were set as covariates in the meta-regression. Funnel plots judged the publication biases, and a P < 0:05 was considered statistically significant [17].

IRB Approval.
This meta-analysis study was approved by the institutional review board of the Department of Hepatobiliary and Pancreas Surgery, The First Affiliated Hospital, Xi'an Jiaotong University.

Study Quality.
The study quality for all 17 independent studies is shown in Table 2.

Meta-Analysis of SSD versus LSD for T2DM.
There was high heterogeneity among the studies, and the random effect model was used. Meta-analysis showed that the incidence of T2DM was not significantly different between SSD and LSD (RR = 0:97, 95% CI: 0.89-1.05, P = 0:41) (Figure 8).

Sensitivity Analyses.
Sensitivity analyses of the association between SSD, LSD, and T2DM were conducted. The results indicated that the sensitivity of the association between SSD and T2DM was robust after each study was excluded one by one. The RR was 1.22 and 95% CI was 1.15-1.29 (Figure 9). The results indicated that the sensitivity of the association between LSD and T2DM was robust after each study was excluded one by one. The RR was 1.26 and 95% CI was 1.15-1.39 ( Figure 10). When the Chinese literature was excluded, the meta-analysis results and heterogeneity did not change significantly. However, according to gender subgroup analysis, the heterogeneity of the male      9 Journal of Diabetes Research 4.6. Publication Bias Analyses. Funnel plots of publication bias for the association between SSD, LSD, and T2DM were assessed. The symmetry found in the funnel plots indicated that the publication bias was weak. (Figures 11 and 12) Most of the studies are at the top of the funnel plot, indicating that the quality of the studies is good.

Discussion
The countries with the largest numbers of adults with DM aged    The risk of T2DM in females with SSD was 1.34 times higher than that with NSD. The risk of T2DM in males with SSD was slightly higher than that in females. Subgroup analysis by region showed that there was no significant difference among the Asian, American, and European subgroups. The risk of T2DM in Asia with SSD was 1.79 times higher than that with NSD. The risk of T2DM in America with SSD was 1.34 times higher than that with NSD. The risk of T2DM in Europe with SSD was 1.24 times higher than that with NSD. The mechanism of SSD leading to T2DM is still unclear. The possible reason is that sleep deprivation causes the imbalance of the sympathetic vagus nerve, which leads to the decrease of β cell response ability, the inhibition of insulin secretion, and the further development of insulin resistance and T2DM. Lack of sleep may also cause the release of a large number of inflammatory factors, inhibits the activity of islet receptor L-arginine kinase, and leads to insulin resistance. Some studies found that slow-wave sleep duration decreased insulin sensitivity and increased the risk of T2DM [46][47][48].
In this meta-analysis of the association between LSD and T2DM, the risk of T2DM with LSD was 1.26 times higher than that with NSD, which was consistent with the conclusions of other cohort studies [37][38][39][40][41][42][43][44][45]. The risk of T2DM in male with LSD was 1.55 times higher than that with NSD. The risk of T2DM in females with LSD was 1.27 times higher than that with NSD. Subgroup analysis by region showed that there were differences among the Asian, American, and European subgroups. The risk of T2DM in Asia with LSD was 1.12 times higher than that with NSD. The risk of T2DM in America with LSD was 1.27 times higher than that with NSD. The risk of T2DM in Europe with LSD was 1.49 times higher than that with NSD. There are few studies on the relationship between LSD and T2DM. The reason may be that long sleeps are a poor sleep quality actually and they may prolong sleep duration to make up for the impact of poor sleep quality. Sleeping for a long time is harmful to health itself. Excessive sleep duration may be associated with other factors, such as obesity, poor social and economic status, low physical activity, depression, sleep apnea, or other chronic diseases. These factors can produce a confounding effect [46,47,49].
The meta-analysis found that LSD or SSD will increase the risk of T2DM. Some studies suggest that LSD or SSD can lead to impaired fasting blood glucose, abnormal HbA1, and insulin resistance [50][51][52][53][54][55][56]. There was a Ushaped dose response relationship between sleep duration and risk of T2DM [10,57]. SSD or LSD increased the risk of T2DM. Therefore, the risk of T2DM can be reduced if the sleep duration of the patients with poor sleep is changed and the sleep duration is maintained between 7 and 8 hours. SSD or LSD is a risk factor for chronic diseases such as DM, coronary heart disease, stroke, and obesity [7,9,58,59]. The sensitivity of each study was robust and the publication bias was weak. While we found that the study location may have been responsible for this heterogeneity in the meta-analysis of the relationship between LSD and T2DM, other factors, such as age, cannot be fully extracted from the text.

Limitations.
Limitations are listed as follows: (1) some studies have not been retrieved. All the included studies are published in Chinese and English, and there may be incomplete literature retrieval, (2) the overall heterogeneity of the study is high, suggesting that there is heterogeneity among the studies. However, after subgroup analyses were performed, the heterogeneity of subgroups decreased significantly, (3) the study explored the association between sleep duration and T2DM but paid less attention to the association between sleep quality and T2DM, which also indicated that we should focus on it in later research, and (4) for special populations such as gestational diabetes, postpartum diabetes, and posttransplant diabetes, there is less attention. More and more studies have confirmed that an unhealthy lifestyle plays an important role in the pathogenesis of T2DM and glucose control in DM [60]. From the perspective of prevention, appropriate sleep duration can be used as the primary prevention of T2DM [7,9,61,62]. Good sleep should be considered as an important health component in the prevention and treatment of T2DM.

Conclusions
The purpose of the current study was to determine the cohort studies of the relationship between sleep duration and T2DM in adults with a systematic review and meta-analysis. This study provides the first systematic assessment of the cohort study of the relationship between sleep duration and T2DM. The findings indicate that SSD or LSD is a risk factor for T2DM. The findings of this study have several important implications for future practice. Further research is required to attend to the association between sleep quality and T2DM.

Data Availability
The data used to support the findings of this study are included within the article.

Ethical Approval
This meta-analysis study was approved by the institutional review board of the Department of Hepatobiliary and Pancreas Surgery, The First Affiliated Hospital, Xi'an Jiaotong University. All protocols conformed with the ethical guidelines of the 1975 Helsinki Declaration, and written, informed consent was obtained from all patients.

Disclosure
None of the authors have any financial and personal relationships with other people or organizations that could inappropriately influence their work.