J-curve relation between daytime nap duration and type 2 diabetes or metabolic syndrome: A dose-response meta-analysis

Adequate sleep is important for good health, but it is not always easy to achieve because of social factors. Daytime napping is widely prevalent around the world. We performed a meta-analysis to investigate the association between napping (or excessive daytime sleepiness: EDS) and the risk of type 2 diabetes or metabolic syndrome, and to quantify the potential dose-response relation using cubic spline models. Electronic databases were searched for articles published up to 2016, with 288,883 Asian and Western subjects. Pooled analysis revealed that a long nap (≥60 min/day) and EDS were each significantly associated with an increased risk of type 2 diabetes versus no nap or no EDS (odds ratio 1.46 (95% CI 1.23–1.74, p < 0.01) for a long nap and 2.00 (1.58–2.53) for EDS). In contrast, a short nap (<60 min/day) was not associated with diabetes (p = 0.75). Dose-response meta-analysis showed a J-curve relation between nap time and the risk of diabetes or metabolic syndrome, with no effect of napping up to about 40 minutes/day, followed by a sharp increase in risk at longer nap times. In summary, longer napping is associated with an increased risk of metabolic disease. Further studies are needed to confirm the benefit of a short nap.

. Search terms used for the electronic databases

Objectives
To investigate the association between napping or excessive daytime sleepiness and the risk of diabetes or metabolic syndrome, and to quantify the potential dose-response relation by using cubic spline models.

Eligibility criteria
Observational studies that reported risk estimates for type 2 diabetes and metabolic syndrome in relation to daytime napping and excessive daytime sleepiness in the general population, and that provided point estimates of odds ratio with the 95% confidence interval or standard error for qualitative assessment.

Study Records
Two authors (TYamada, NS) will independently perform the searches. Literature search results will be uploaded to EndNote Data management Jan 2016 11a Two authors (TYamada, NS) will independently screen titles/abstracts and obtain full reports for 1) reports meeting inclusion criteria; 2) those requiring further discussion. Any discrepancies will be resolved through discussion.

Selection Process Jan 2016 11b
Extracted data will be independently (TYamada, NS) add into digital pre-defined forms (Excel). Data collection process Jan 2016 11c

Data Items
We extract information on the characteristics of each study (study name, authors, year of publication, journal, study type, study location, and number of participants and incident cases), the subject characteristics (age, sex, and BMI), the extent of exposure to napping (definition of napping, nap time, and prevalence of napping in each category), the validity of the method used for assessment of napping (and excessive daytime sleepiness), the validity of the method used for assessment of the outcome (diabetes and metabolic syndrome), and the validity of the analytical methods (statistical models, covariates included in the models, and risk estimates for each nap duration category).

Outcomes and Prioritization
The odds ratio (OR) and its 95% confidence interval (CI) will be employed as the measure of association in all studies -Jan 2016 13

Risk of Bias in Individual Studies
Study-level quality will be assessed using the Newcastle Ottawa Scale by two authors (TYamada, NS) and disagreement will be resolved through discussion.

Data Synthesis
We will conduct a meta-analysis for each outcome using the DerSimonian-Laird random effects model to compare napping categories and set study weights as equal to the inverse variance of the estimated effect for each study. To evaluate the potential dose-response relation between diabetes and nap time, a dose-response meta-analysis will be performed taking into account the between-study heterogeneity proposed by Orsini et al. to compute the trend from correlated log values of OR estimates across various nap times. A restricted cubic spline model for the duration of nap time with three knots (5th, 35th, 65th, and 95th percentiles) will be estimated by generalized least squares regression analysis, taking into account the correlations within each set of published ORs. Probability (P) values for curve linearity or nonlinearity will be calculated by testing the null hypothesis that the coefficient of the second spline equals zero. This analysis will incorporate data on the ORs and 95% CIs, the number of cases and participants, and the median or mean nap time (minutes per day) for each group.
The midpoint of the upper and lower borders will be set as the median dose for each category if the median or mean exposure per category was not reported. If the highest category is open-ended, the midpoint of the category was set at 1.25 times the lower border. For the lowest nap category, we set the median at 0.5 times the cut-off point (e.g., if category was <30 min, the median was set at 15 min). -

Jan 2016 15a
Cochrane's I2 test and the I2 test will be used to evaluate heterogeneity among the studies. Stratified analyses will be also performed (with stratification by study location, study score, and study type). Possible publication bias will be evaluated by creating a funnel plot of the effect size for each study versus the standard error. Then asymmetry of the funnel plots will be was assessed by performing Begg's test and Egger's test.

Meta-bias(es)
Tables will show the availability of data for each study and outcome (selective reporting) -Jan 2016 16

Confidence in cumulative evidence
Results will be commented in view of study limitations and available evidence Participants were asked to report whether they were taking a midday nap, as well as its frequency and duration.
Participants were classified as having diabetes if they reported a history of diagnosis of the disease, took antidiabetic drugs, or had a non-fasting glucose level ≥200 mg/dl.
Xu et al, 2010 Participants were asked to report whether they were taking a midday nap, as well as its frequency and duration.
Participants were asked whether they had ever been told that they had diabetes by a doctor.
Fang et al, 2013 Participants were asked to report whether they were taking a midday nap, as well as its frequency and duration.
Participants were classified as having diabetes if they reported a history of diagnosis of diabetes by a physician, were on antidiabetic medication, or had a high fasting plasma glucose level (≥7.0 mmol/L).
Lam et al, 2010 Participants were asked to report whether they were taking a midday nap, as well as its frequency and duration.
Participants were classified as having diabetes if they reported a history of diagnosis of diabetes by a physician, were on antidiabetic medication, or had a high fasting plasma glucose level (≥7.0 mmol/L).

Author, Year
Definition of EDS Definition of diabetes Lindberg et al. 2007 (21) Participants were asked to report whether they fell asleep involuntarily for a short period during the day, e.g., when there was a pause at work.
Participants were asked whether they had ever been told that they had diabetes by a doctor. Bixer et al. 2005 (22) Participants were asked to report whether they felt drowsy or sleepy most of the day, but manage to stay awake.
Participants were classified as having diabetes if they had current treatment for diabetes or a fasting blood glucose level >126 mg/dl. Renko et al. 2005 (23) Participants were asked to report whether they noted sleepiness in the daytime.
Participants were classified as having diabetes if they fulfilled the criteria for diabetes in the75g OGTT. Asplund. 1995 (24) Participants were asked to report whether they were often sleepy in the daytime.
Participants were asked whether they had ever been told that they had diabetes by a doctor.