Social jet lag and (changes in) glycemic and metabolic control in people with type 2 diabetes
Funding information: Dutch Diabetes Foundation Diabetes Fonds Fellowship, Grant/Award Number: 2019.82.002
Abstract
Objective
Social jet lag, i.e., the discordance among social and biological rhythms, is associated with poor metabolic control. This study aimed to assess cross-sectional and longitudinal associations among social jet lag and glycemic and metabolic control in people with type 2 diabetes.
Methods
In a prospective cohort (N = 990) with type 2 diabetes, social jet lag was measured at baseline using daily diaries and was categorized (high, moderate, or low). Metabolic outcomes were assessed at baseline and at 1 and 2 years of follow-up. Associations among social jet lag and glycemic and metabolic control were analyzed using linear regression and linear mixed models adjusted for confounding factors. Analyses were stratified for work status (retired vs. working; p value for interaction = 0.007 for glycated hemoglobin [HbA1c]).
Results
In working people, a cross-sectional association between high social jet lag and HbA1c (1.87 mmol/mol [95% CI: 0.75 to 2.99]) and blood pressure (5.81 mm Hg [95% CI: 4.04 to 7.59]) was observed. For retired people, high social jet lag was negatively associated with HbA1c (−1.58 mmol/mol [95% CI: −2.54 to −0.62]), glucose (−0.19 mmoL/L [95% CI:−0.36 to −0.01]), and blood pressure (−3.70 mm Hg [95% CI: −5.36 to −2.04]), and the association with BMI was positive (1.12 kg/m2 [95% CI: 0.74 to 1.51]). Prospective associations had the same direction as cross-sectional findings but were nonsignificant for working or retired people.
Conclusions
Social jet lag was cross-sectionally, but not prospectively, associated with glycemic and metabolic markers. Interaction with work status was present, and directions of the associations were generally detrimental in the working population, whereas higher social jet lag was associated with improved glycemic and metabolic control for retired people.
Study Importance
What is already known?
- Social jet lag is a subtle but chronic disruption of sleep timing that is highly prevalent.
- It has been associated with poor metabolic and glycemic control.
What does this study add?
- Our results show that social jet lag is cross-sectionally, but not prospectively, associated with glycemic and metabolic markers.
How might these results change the direction of research?
- Future studies should focus on the longitudinal effects of social jet lag on glycemic and metabolic control and explore the mechanisms and pathways that are involved.
INTRODUCTION
Recently, the body's circadian clock has gained interest as a risk factor for type 2 diabetes (T2D) [(1-3)]. The circadian clock aligns physiological and behavioral processes in the body with the environment. Disturbance in the synchronization of this circadian rhythm can have negative effects on these processes [(4)]. Shift work, for example, causes circadian disruptions, which have been associated with impaired glucose and lipid metabolism, increased risk of obesity, and a higher prevalence of T2D and cardiovascular diseases [(4)].
Social jet lag is a circadian disruption that is much more subtle, but chronic. It is described as a disagreement among the biological rhythm, determined by the circadian clock, and the sleep and wake-up times imposed by work schedules and social engagements [(5)]. Social jet lag is demonstrated by different sleep timing during workdays compared with free weekend days, when people tend to go to bed later and sleep in in the morning. It is measured as the difference in midpoint of sleep between week and weekend days. With a prevalence above 50% in the general population, social jet lag is a highly prevalent circadian disruptor [(5, 6)].
Our recent meta-analysis of the associations among social jet lag and parameters of the metabolic syndrome and T2D has shown that social jet lag is associated with increased levels of glycated hemoglobin (HbA1c), higher body mass index (BMI), higher waist circumference, and higher odds of obesity [(7)]. Social jet lag has also been associated with worse glycemic control, higher HbA1c levels, and higher insulin requirements in people with type 1 diabetes [(8-11)]. Few studies have assessed the effects of social jet lag in people with T2D. These studies have suggested that sleep disruptions and social jet lag impact glycemic control in people with T2D [(12-15)]. For instance, Kelly and colleagues found that social jet lag was significantly associated with higher HbA1c levels in a group of 252 people with T2D [(13)]. However, all previously performed studies are limited because of cross-sectional study designs [(12-15)], small sample sizes [(13, 14)], or inadequate correction for possible confounders such as sex, age, education, work status, and diabetes complications [(15)]. For example, the prevalence of social jet lag is highly influenced by age [(16, 17)], and age also has been strongly associated with metabolic and glycemic markers [(5)]. In our population, this also holds true for diabetic complications such as kidney function (estimated glomerular filtration rate) and neuropathy. These have been associated with sleep, and in addition, they have been strongly associated with our metabolic and glycemic outcomes [(18-21)]. Without taking these confounding factors into account, the associations among social jet lag and metabolic and glycemic markers could be over- or underestimated.
Therefore, we aimed to investigate the longitudinal associations among social jet lag and (changes in) glycemic and metabolic control in a cohort of people with T2D over 2 years of follow-up.
METHODS
Study design and population
We used data from the Diabetes Care System cohort, a prospective cohort of people with T2D from the West Friesland region of the Netherlands [(22)]. Inclusion in the Diabetes Care System cohort started in 1998, and, currently, the cohort comprises around 15,000 people with T2D. Participants visit the care center annually for monitoring of their disease and comorbidities [(22)]. Between June 2017 and July 2018, a total of 3592 people from this cohort were approached for participation in a one-time substudy on lifestyle and environmental factors, in addition to their regular visits. A total of 1549 people participated in this substudy (Figure 1). Informed consent was obtained from all participants, and the study was approved by the ethical committee of the Vrije Universiteit Medical Centre (UMC).
From the total of 1549 participants at baseline, 29 were excluded because they did not have established T2D, and 530 participants were excluded because of missing information on social jet lag (Figure 1). These participants had similar characteristics as the included participants (male sex: 60.1%, age: 68.4 ± 9.8 years, HbA1c: 55.2 ± 13.0 mmol/mol). The final analytic sample for this study consisted of 990 participants (Figure 1). For this substudy, we collected lifestyle and environmental data. For this analysis, we combined this substudy data with annual routine care data at baseline, and, after 1 and 2 years, this included blood glucose, HbA1c, BMI, blood pressure, and blood lipids [(22)].
Social jet lag measurements
Sleep duration was measured for 1 week during week and weekend days using a sleep diary. Participants were asked to report sleeping hours in their diary every day. Social jet lag was determined as the difference in midpoint of sleep in hours between week days and weekend days [(6)]. Subsequently, social jet lag was calculated as a continuous variable, in hours and minutes, and categorized into three groups based on literature as low social jet lag (≤1 hour), moderate social jet lag (>1 hour and <2 hours), and high social jet lag (≥2 hours) [(6, 23)].
Glycemic and metabolic measurements
Glucose levels were determined at baseline and at the two following annual care visits. Blood samples were collected after an overnight fast. HbA1c levels were determined with turbidimetric inhibition immunoassay for hemolyzed whole EDTA blood (Cobas c501, Roche, Basel, Switzerland) and were expressed as percentage and in millimoles per mole [(22)]. Fasting plasma glucose (FPG) was assessed in fluorinated plasma with the ultraviolet exposure (UV) test using hexokinase (Cobas c501) [(22)].
Total cholesterol, high-density lipoprotein (HDL) cholesterol, low-density lipoprotein (LDL) cholesterol, and triglyceride levels were assessed at baseline and at the two follow-up visits. Total cholesterol, HDL cholesterol, and triglycerides were determined enzymatically (Cobas c501) from blood samples obtained as described earlier [(22)]. LDL cholesterol concentration was calculated as follows: LDL cholesterol = cholesterol − HDL cholesterol – 0.45 × triglycerides [(22)].
Systolic and diastolic blood pressure was measured at baseline and at the two follow-up visits. Measurements were performed after 5 minutes of rest in a seated position on the right arm using an oscillometric device (Welch Allyn ProBP 3400, Skaneateles Falls, New York) and repeated after 3 minutes of rest [(22)].
At the baseline annual visit and at the two follow-up visits, weight (kilograms) and height (centimeters) were measured, with people wearing only light clothes and barefoot, and BMI was calculated as weight in kilograms divided by height in meters squared.
Covariates
Medical records were used to determine age (years), sex (male or female), education (low, middle, or high), and diabetes duration (years). Work status was self-reported as currently working, volunteer working, unemployed, or retired. We defined work status as retired versus not retired. Data on shift work were collected in a subsample of 40% of the population. None of those 40% worked evening or night shifts, suggesting a very low prevalence of people working shifts. During the baseline annual visits, the presence of neuropathy was diagnosed based on the loss of protective sensibility and/or peripheral arterial disease or history of ulcer or amputation, determined with a modified Simm's classification [(22)]. Kidney function was determined with estimated glomerular filtration rate, corrected for age, sex, and creatinine levels [(24)]. Average sleep duration was determined from sleep diaries.
Additionally, use of oral glucose-lowering medication and insulin use were assessed and defined using Anatomical Therapeutic Chemical (ATC) classification codes A10B and A10A, respectively. The use of diabetes medication was categorized as use of one oral glucose-lowering drug, two oral glucose-lowering drugs, and more than two oral glucose-lowering drugs or use of insulin. Use of one or more lipid-lowering drugs was defined using ATC codes A10, and use of one or more antihypertensive drugs was defined by combining the ATC codes C02, C03, C04, C07, C08, and C09. Change in the use of these medications was defined as an increase, decrease, or stable status over 2 years of follow-up.
Statistical analysis
Statistical analysis was performed using RStudio version 4.0.3 (R, Vienna, Austria). Variables were visually assessed for normal distribution using quantile-quantile (Q–Q) plots. Descriptive data are expressed as means ± standard deviation or medians (interquartile range) for numeric variables, depending on their distribution, and as numbers (percentage) for categorical variables. Baseline descriptive characteristics are reported stratified for three categories of social jet lag.
To address missing values, multiple imputation was used by predictive mean matching, logistic regression, and poly regression, as appropriate. Work status had the highest amount of missing values (8.9%). All covariates with missing values were imputed. Ten imputation sets were performed with 50 iterations, and results were gained from pooled estimates.
Effect modification was tested for variables based on literature. Age and work status were added to the regression models by including their interaction terms. Because we have lower power in interaction analysis, effect modification was considered significant from p ≤ 0.1 [(25)].
Cross-sectional analysis
We used linear regression models to assess the cross-sectional associations among social jet lag and several parameters of glycemic and metabolic control (HbA1c, FPG, cholesterol ratio, systolic blood pressure, and BMI). The regression models were corrected for several confounders based on literature. Model 1 was corrected for sex, age, and education. Model 2 was additionally corrected for diabetes duration, complications, and medication use. Model 3 was additionally corrected for sleep duration. We did not adjust for lifestyle factors because of their possible mediating role in the association.
Prospective analysis
We used linear mixed regression models to assess the prospective associations among social jet lag and parameters of glycemic and metabolic control (HbA1c, FPG, cholesterol ratio, systolic blood pressure, and BMI) at baseline and at the two follow-up measurements. The same three confounder-corrected models are presented for the cross-sectional associations, with model 2 and 3 now corrected for change in medication use over the follow-up time.
A two-sided p ≤ 0.05 was considered statistically significant. Unstandardized regression coefficients, 95% confidence intervals (CI), and p values are presented.
RESULTS
Population characteristics
In our cohort, 346 participants had low social jet lag (≤1 hour social jet lag), 435 participants were categorized as having moderate social jet lag (>1 hour and <2 hours social jet lag), and 209 participants had high social jet lag (≥2 hours social jet lag; Table 1). Participants with ≥2 hours of social jet lag, in comparison with participants with 1 hour less of social jet lag, were younger (62.2 vs. 72.3 years), had a shorter diabetes duration (11.0 vs. 11.9 years), and had neuropathy less often (21.7% vs. 31.8%). Moreover, participants with high social jet lag were most often working (paid jobs; 54.5%), whereas participants with low and moderate social jet lag were mostly retired (7.9% and 20.3% working, respectively). Total sleep duration was similar among all groups.
| ≤1 Hour (n = 346) | >1 Hour and <2 Hours (n = 435) | ≥2 Hours (n = 209) | |
|---|---|---|---|
| Social jet lag (h:mm) | 0:38 ± 0:13 | 1:24 ± 0:18 | 2:55 ± 0:53 |
| Sleep duration (h:mm; n = 775) | 8:38 ± 0:54 | 8:33 ± 0:57 | 8:14 ± 1:10 |
| Male sex (%) | 210 (60.7) | 270 (62.1) | 157 (75.1) |
| Age (y) | 72.3 ± 7.2 | 68.5 ± 9.1 | 62.2 ± 9.0 |
| BMI (kg/m2; n = 988) | 28.8 ± 4.9 | 29.7 ± 4.9 | 30.1 ± 5.0 |
| Glycemic and metabolic | |||
| Diabetes duration (y) | 11.9 (8.1; 16.8) | 11.2 (8.2; 16.0) | 11.0 (8.1; 15.6) |
| HbA1c (mmol/mol; n = 989) | 53.1 ± 11.0 | 54.0 ± 13.5 | 54.3 ± 13.0 |
| Fasting glucose (mmol/L; n = 972) | 8.5 ± 2.0 | 8.7 ± 2.3 | 8.9 ± 2.5 |
| Neuropathy (%; n = 958) | 107 (31.8) | 117 (27.9) | 44 (21.7) |
| Corrected eGFR1 (mL/min) | 71.8 ± 16.5 | 73.9 ± 18.8 | 79.9 ± 20.4 |
| Oral medication or insulin (%) | |||
| No medication | 61 (17.6) | 67 (15.4) | 36 (17.2) |
| 1 Oral medication | 106 (30.6) | 141 (32.4) | 63 (30.1) |
| 2 Oral medications | 79 (22.8) | 104 (23.9) | 50 (23.9) |
| Insulin | 99 (28.6) | 121 (27.8) | 58 (27.8) |
| >2 Oral medications | 1 (0.3) | 2 (0.5) | 2 (1.0) |
| Antihypertensive medication | 279 (80.6) | 329 (75.6) | 148 (70.8) |
| Lipid-lowering medication | 279 (80.6) | 337 (77.5) | 164 (78.5) |
| HDL cholesterol (mmol/L; n = 989) | 1.3 ± 0.4 | 1.3 ± 0.4 | 1.2 ± 0.4 |
| LDL cholesterol (mmol/L) | 2.2 ± 0.8 | 2.2 ± 0.9 | 2.2 ± 0.8 |
| Total cholesterol (mmol/L) | 4.2 ± 1.0 | 4.2 ± 1.1 | 4.1 ± 1.0 |
| Systolic blood pressure (mm Hg; n = 989) | 141.9 ± 19.5 | 141.2 ± 21.1 | 135.9 ± 16.6 |
| Diastolic blood pressure (mm Hg; n = 989) | 77.5 ± 8.5 | 878.0 ± 8.3 | 97.1 ± 7.5 |
| Socioeconomic factors | |||
| Educational level (%; n = 973) | |||
| Low | 108 (31.5) | 117 (27.4) | 53 (26.1) |
| Middle | 174 (50.3) | 197 (46.1) | 100 (49.3) |
| High | 61 (17.8) | 110 (25.8) | 50 (24.6) |
| Other or unknown | NA | 3 (0.7) | NA |
| Work status (%; n = 901) | |||
| Yes | 25 (7.9) | 81 (20.3) | 102 (54.5) |
| Yes, voluntary | 20 (6.3) | 44 (11.0) | 18 (9.6) |
| No | 15 (4.8) | 25 (6.3) | 11 (5.9) |
| No, retired | 255 (81.0) | 249 (62.4) | 56 (29.9) |
- Note: Data are shown as percentages, means ± standard deviation, or medians (interquartile range), depending on the distributions.
- Abbreviations: HbA1c, glycated hemoglobin; HDL, high-density lipoprotein; eGFR, estimated glomerular filtration rate; eGFR1, corrected for sex, age, and creatinine levels; LDL, low-density lipoprotein; NA, not available; T2D, type 2 diabetes.
Glycemic control
We observed effect modification by work status (for example, p value for interaction = 0.007 for HbA1c) and, therefore, stratified the analyses for retired versus working people. Cross-sectionally (Table 2), we found that high and moderate social jet lag was significantly associated with higher HbA1c levels compared with low social jet lag (1.25 mmol/mol [95% CI: 0.21 to 2.30] and 1.87 mmol/mol [95% CI: 0.75 to 2.99]), respectively. In retired people, moderate compared with low social jet lag was significantly associated with higher HbA1c levels (0.72 mmol/mol [95% CI: 0.13 to 1.31]), whereas high compared with low social jet lag was significantly associated with lower HbA1c levels (−1.58 mmol/mol [95% CI: −2.54 to −0.62]). In working people, social jet lag was not significantly associated with FPG levels. However, in retired people, moderate compared with low social jet lag was associated with 0.15 mmoL/L (95% CI: 0.05 to 0.26) higher FPG, whereas high compared with low social jet lag was associated with −0.19 mmoL/L lower FPG levels (95% CI: −0.36 to −0.01).
| Working | ||||
|---|---|---|---|---|
| Outcome | Model 1 | Model 2 | Model 3 | |
| HbA1c | ≤1 h | Reference | Reference | Reference |
| >1 h and <2 h | 1.24 (0.06 to 2.42) | 1.30 (0.26 to 2.34) | 1.25 (0.21 to 2.30) | |
| ≥2 h | 1.81 (0.56 to 3.06) | 2.01 (0.90 to 3.12) | 1.87 (0.75 to 2.99) | |
| Fasting glucose | ≤1 h | Reference | Reference | Reference |
| >1 h and <2 h | −0.09 (−0.31 to 0.14) | −0.11 (−0.32 to 0.10) | −0.10 (−0.31 to 0.11) | |
| ≥2 h | 0.13 (−0.11 to 0.37) | 0.14 (−0.09 to 0.36) | 0.17 (−0.06 to 0.39) | |
| Retired | ||||
|---|---|---|---|---|
| Outcome | Model 1 | Model 2 | Model 3 | |
| HbA1c | ≤1 h | Reference | Reference | Reference |
| >1 h and <2 h | 0.73 (0.08 to 1.37) | 0.73 (0.14 to 1.31) | 0.72 (0.13 to 1.31) | |
| ≥2 h | −2.24 (−3.30 to −1.18) | −1.55 (−2.51 to −0.59) | −1.58 (−2.54 to −0.62) | |
| Fasting glucose | ≤1 h | Reference | Reference | Reference |
| >1 h and <2 h | 0.15 (0.04 to 0.26) | 0.15 (0.04 to 0.25) | 0.15 (0.05 to 0.26) | |
| ≥2 h | −0.26 (−0.44 to −0.08) | −0.19 (−0.37 to −0.02) | −0.19 (−0.36 to −0.01) | |
- Note: Statistically significant findings are in bold. Units: HbA1c in millimoles per mole; fasting glucose in millimoles per liter. Model 1 = corrected for sex, age, and education level; model 2 = model 1 additionally corrected for diabetes duration, neuropathy, estimated glomerular filtration rate, use of glucose-lowering drugs, use of lipid-lowering drugs, and use of antihypertensive drugs; and model 3 = model 2 additionally corrected for average sleep duration.
- Abbreviations: HbA1c, glycated hemoglobin; T2D, type 2 diabetes.
Prospectively, over 2 years of follow-up (Table 3), we found no statistically significant associations among social jet lag and HbA1c and fasting glucose levels. However, for both HbA1c and fasting glucose, we observed a similar direction of estimates as the cross-sectional analyses, including the differences among working and retired people.
| Working | ||||
|---|---|---|---|---|
| Outcome | Model 1 | Model 2 | Model 3 | |
| HbA1c | ≤1 h | Reference | Reference | Reference |
| >1 h and <2 h | 0.83 (−2.09 to 3.75) | 0.72 (−1.90 to 3.33) | 0.70 (−1.91 to 3.32) | |
| ≥2 h | 1.63 (−1.52 to 4.78) | 1.76 (−1.07 to 4.59) | 1.72 (−1.12 to 4.57) | |
| Fasting glucose | ≤1 h | Reference | Reference | Reference |
| >1 h and <2 h | 0.03 (−0.48 to 0.54) | 0.02 (−0.45 to 0.48) | 0.02 (−0.45 to 0.48) | |
| ≥2 h | 0.07 (−0.48 to 0.62) | 0.09 (−0.41 to 0.59) | 0.09 (−0.41 to 0.60) | |
| Retired | ||||
|---|---|---|---|---|
| Outcome | Model 1 | Model 2 | Model 3 | |
| HbA1c | ≤1 h | Reference | Reference | Reference |
| >1 h and <2 h | 1.01 (−0.56 to 2.58) | 0.98 (−0.42 to 2.37) | 0.98 (−0.42 to 2.38) | |
| ≥2 h | −0.93 (−3.42 to 1.57) | −0.24 (−2.46 to 1.99) | −0.23 (−2.45 to 2.00) | |
| Fasting glucose | ≤1 h | Reference | Reference | Reference |
| >1 h to <2 h | 0.15 (−0.12 to 0.42) | 0.13 (−0.13 to 0.39) | 0.13 (−0.12 to 0.39) | |
| ≥2 h | −0.23 (−0.66 to 0.20) | −0.17 (−0.58 to 0.24) | −0.15 (−0.56 to 0.26) | |
- Note: Units: HbA1c in millimoles per mole; fasting glucose in millimoles per liter. Model 1 = corrected for sex, age, and education level; model 2 = model 1 additionally corrected for diabetes duration, neuropathy, estimated glomerular filtration rate, use of glucose-lowering drugs, use of lipid-lowering drugs, and use of antihypertensive drugs; and model 3 = model 2 additionally corrected for average sleep duration.
- Abbreviations: HbA1c, glycated hemoglobin; T2D, type 2 diabetes.
Metabolic control
Cross-sectionally (Table 4), for working people, having moderate and high social jet lag was significantly associated with 7.86 mm Hg (95% CI: 6.21 to 9.52) and 5.81 mm Hg (95% CI: 4.04 to 7.59) higher systolic blood pressure, respectively, compared with having low social jet lag. For retired people, the cross-sectional association reversed to −0.45 mm Hg (95% CI: −1.46 to 0.56) for moderate social jet lag and −3.70 mm Hg (95% CI: −5.36 to −2.04) for high compared with low social jet lag. For cholesterol ratio, we found a significant negative association for high compared with low social jet lag in working people (−0.18 [95% CI: −0.31 to −0.06]). However, for retired people, we found a significant positive association for moderate compared with low social jet lag (0.17 [95% CI: 0.11 to 0.23]). BMI was not associated with social jet lag in working people but was significantly associated with moderate and high social jet lag in retired people (0.48 kg/m2 [95% CI: 0.25 to 0.71] and 1.12 kg/m2 [95% CI: 0.74 to 1.51], respectively).
| Working | ||||
|---|---|---|---|---|
| Outcome | Model 1 | Model 2 | Model 3 | |
| Systolic blood pressure | ≤1 h | Reference | Reference | Reference |
| >1 h and <2 h | 8.23 (6.57 to 9.90) | 7.81 (6.16 to 9.47) | 7.86 (6.21 to 9.52) | |
| ≥2 h | 6.23 (4.46 to 8.00) | 5.65 (3.89 to 7.42) | 5.81 (4.04 to 7.59) | |
| Cholesterol ratio | ≤1 h | Reference | Reference | Reference |
| >1 h and <2 h | −0.01 (−0.13 to 0.11) | −0.05 (−0.16 to 0.07) | −0.04 (−0.16 to 0.08) | |
| ≥2 h | −0.21 (−0.33 to −0.08) | −0.20 (−0.32 to −0.07) | −0.18 (−0.31 to −0.06) | |
| BMI | ≤1 h | Reference | Reference | Reference |
| >1 h and <2 h | 0.30 (−0.16 to 0.76) | 0.19 (−0.26 to 0.65) | 0.18 (−0.28 to 0.63) | |
| ≥2 h | −0.13 (−0.62 to 0.36) | −0.22 (−0.70 to 0.27) | −0.26 (−0.75 to 0.23) | |
| Retired | ||||
|---|---|---|---|---|
| Outcome | Model 1 | Model 2 | Model 3 | |
| Systolic blood pressure | ≤1 h | Reference | Reference | Reference |
| >1 h and <2 h | −0.58 (−1.61 to 0.44) | −0.45 (−1.46 to 0.56) | −0.45 (−1.46 to 0.56) | |
| ≥2 h | −4.11 (−5.78 to −2.43) | −3.69 (−5.34 to −2.03) | −3.70 (−5.36 to −2.04) | |
| Cholesterol ratio | ≤1 h | Reference | Reference | Reference |
| >1 h and <2 h | 0.15 (0.08 to 0.21) | 0.17 (0.11 to 0.23) | 0.17 (0.11 to 0.23) | |
| ≥2 h | −0.07 (−0.17 to 0.04) | −0.06 (−0.16 to 0.04) | −0.06 (−0.16 to 0.04) | |
| BMI | ≤1 h | Reference | Reference | Reference |
| >1 h and <2 h | 0.45 (0.21 to 0.69) | 0.50 (0.26 to 0.73) | 0.48 (0.25 to 0.71) | |
| ≥2 h | 1.22 (0.82 to 1.61) | 1.18 (0.80 to 1.57) | 1.12 (0.74 to 1.51) | |
- Note: Statistically significant findings in bold. Units: systolic blood pressure in millimeters of mercury; BMI in kilograms per meters squared. Model 1 = corrected for sex, age, and education level; model 2 = model 1 additionally corrected for diabetes duration, neuropathy, estimated glomerular filtration rate, use of glucose-lowering drugs, use of lipid-lowering drugs, and use of antihypertensive drugs; and model 3 = model 2 additionally corrected for average sleep duration.
- Abbreviation: T2D, type 2 diabetes.
Prospectively (Table 5), for systolic blood pressure and cholesterol ratio, we found results in the same direction for working and retired people; however, they were not statistically significant. For BMI, prospectively, no associations were found in working people, whereas retired people with high compared with low social jet lag also had higher BMI (1.13 kg/m2 [95% CI: 0.11 to 2.16]).
| Working | ||||
|---|---|---|---|---|
| Outcome | Model 1 | Model 2 | Model 3 | |
| Systolic blood pressure | ≤1 h | Reference | Reference | Reference |
| >1 h and <2 h | 3.92 (−0.04 to 7.87) | 3.51 (−0.43 to 7.45) | 3.50 (−0.45 to 7.45) | |
| ≥2 h | 2.34 (−1.84 to 6.70) | 1.87 (−2.40 to 6.15) | 1.84 (−2.45 to 6.13) | |
| Cholesterol ratio | ≤1 h | Reference | Reference | Reference |
| >1 h and <2 h | 0.08 (−0.27 to 0.42) | 0.07 (−0.27 to 0.42) | 0.08 (−0.27 to 0.42) | |
| ≥2 h | −0.16 (−0.54 to 0.21) | −0.12 (−0.50 to 0.25) | −0.11 (−0.48 to 0.27) | |
| BMI | ≤1 h | Reference | Reference | Reference |
| >1 h and <2 h | 0.07 (−1.13 to 1.26) | 0.00 (−1.19 to 1.19) | 0.01 (−1.19 to 1.21) | |
| ≥2 h | −0.18 (−1.47 to 1.12) | −0.23 (−1.52 to 1.07) | −0.20 (−1.51 to 1.10) | |
| Retired | ||||
|---|---|---|---|---|
| Outcome | Model 1 | Model 2 | Model 3 | |
| Systolic blood pressure | ≤1 h | Reference | Reference | Reference |
| >1 h and <2 h | 1.33 (−1.20 to 3.85) | 1.43 (−1.08 to 3.94) | 1.41 (−1.11 to 3.92) | |
| ≥2 h | −1.77 (−5.78 to 2.24) | −1.67 (−5.67 to 2.33) | −1.73 (−5.74 to 2.28) | |
| Cholesterol ratio | ≤1 h | Reference | Reference | Reference |
| >1 h and <2 h | 0.16 (−0.02 to 0.34) | 0.17 (0.00 to 0.34) | 0.17 (0.00 to 0.34) | |
| ≥2 h | −0.09 (−0.37 to 0.20) | −0.09 (−0.36 to 0.18) | −0.09 (−0.36 to 0.19) | |
| BMI | ≤1 h | Reference | Reference | Reference |
| >1 h and <2 h | 0.43 (−0.23 to 1.08) | 0.48 (−0.17 to 1.12) | 0.46 (−0.18 to 1.10) | |
| ≥2 h | 1.19 (0.15 to 2.24) | 1.18 (0.16 to 2.21) | 1.13 (0.11 to 2.16) | |
- Note: Statistically significant findings in bold. Units: systolic blood pressure in millimeters of mercury; BMI in kilograms per meters squared. Model 1 = corrected for sex, age, and education level; model 2 = model 1 additionally corrected for diabetes duration, neuropathy, estimated glomerular filtration rate, use of glucose-lowering drugs, use of lipid-lowering drugs, and use of antihypertensive drugs; and model 3 = model 2 additionally corrected for average sleep duration.
- Abbreviation: T2D, type 2 diabetes.
DISCUSSION
In this study, we aimed to investigate the associations among social jet lag and (changes in) glycemic and metabolic control in a cohort of people with T2D over 2 years of follow-up. Social jet lag was present in 65% of our population. We observed effect modification by work status (p < 0.1 for 60% of the outcomes). For working people, we found cross-sectional associations among moderate and high social jet lag and higher HbA1c and systolic blood pressure. No associations were found for FPG and BMI. Prospective analyses confirmed these findings; however, they were not significant. For retired people, we found contrasting results, with protective associations for high social jet lag with HbA1c, glucose, and systolic blood pressure but also significant associations with increased cholesterol and BMI. Again, prospective analyses, albeit not significant, confirmed these findings.
Overall, our results show that social jet lag is associated with several glycemic and metabolic markers. Previous research has supported these findings, as our recent meta-analysis showed that social jet lag was significantly associated with higher BMI, waist circumference, HbA1c levels, and systolic blood pressure in general but also in people with diabetes [(7)]. This is also in agreement with findings in people with type 1 diabetes [(8-11)]. In this study, however, we only found cross-sectional associations among social jet lag and markers of glycemic and metabolic control, and not prospectively.
This lack of prospective association might be explained by the fact that our study population consisted of well-treated people with T2D. The participants in the Diabetes Care System cohort are under strict annual monitoring of their diabetes and comorbidities at the centralized diabetes care center, which results in relatively healthy people with diabetes compared with people who receive general diabetes care [(26)]. For example, the average HbA1c level in our population is 53.8 mmol/mol (7.1%) compared with 63.0 mmol/mol (7.9%) in a United States population with the same diabetes duration [(27)]. Over 2 years of follow-up, our study population increased HbA1c only by 3.6 mmol/mol (0.3%). With these small changes, it is also unlikely we could show differences in glycemic and metabolic control due to social jet lag. A larger sample of participants from this cohort and a longer follow-up period might be needed to find possible predictors of change over time.
A final finding that needs to be discussed is the moderating effect of work status in this cohort. We observed differences in the associations among social jet lag and glycemic and metabolic outcomes in people who were still working versus retired people. This difference has also been shown in other studies [(6, 16)], and it could be caused by a difference in age and lifestyle. For the elderly and retired population, social jet lag could be an indication of maintaining an active social life and large social network. These are factors that are known to have many beneficial health effects [(28)]. Therefore, social jet lag might be a different concept for working and retired people, implicating different consequences. This advocates for the need to stratify for age or work status in social jet lag research.
Mechanisms
There are several possible mechanisms suggested in the literature that might play a role in the detrimental effects of social jet lag. Previous research has shown that lifestyle factors are associated to glycemic and metabolic outcomes [(29)]. Additionally, social jet lag causes circadian misalignment, which is associated with disruption of the hypothalamic-pituitary-adrenal axis, which promotes metabolic and glycemic changes [(30, 31)]. Furthermore, sleep architecture, mood, incretin hormones, and visceral fat distribution are suggested to be involved [(32-36)]. In the associations with blood pressure, specifically, stress or job strain might also play a role [(37, 38)]. A final factor could be timing of medication administration. Several studies have suggested that blood pressure medication is more effective when taken daily, at the same time, and before patients go to bed [(39, 40)]. Especially in people with T2D, bedtime hypertension treatment is associated with improved systolic and diastolic blood pressure control and lower prevalence of cardiovascular disease risk factors such as glucose and cholesterol levels [(41)]. Variability in sleep timing such as social jet lag might thereby interfere with the effectiveness of the medication. This is an especially relevant pathway for our study population because a large proportion (>75%) uses blood pressure medication. It is crucial to investigate these pathways further in the future to assess how exactly social jet lag affects glycemic and metabolic control.
Strengths and limitations
Strengths of this study are its prospective design and follow-up period of 2 years. Additionally, in contrast to other studies, we measured many covariates and were thereby able to adjust for a variety of confounders. Moreover, we measured several outcome parameters related to glycemic and metabolic control to assess the effect of social jet lag more broadly.
Our study also has limitations. We have assessed social jet lag only at baseline, even though it is possible to change over time. Therefore, it would have been valuable to assess social jet lag at multiple time points. Although we had a large study population, the largest part of our study population was retired (61.9%), whereas we see most negative effects of social jet lag in working people in the ages between 30 and 60 years. A larger study population in that range would have more power and could possibly identify more associations among social jet lag and glycemic and metabolic control. Furthermore, as mentioned earlier, our study population receives annual diabetes monitoring and centralized care. This approach has been shown to result in better patient outcomes [(26)]. Therefore, it is likely that our study population is healthier compared with other countries or care systems. This implies that our results might not translate very well to a general diabetes population. Finally, because changes in glycemic and metabolic outcomes develop gradually, a longer follow-up period, e.g., 5 years, could be more suitable.
Future perspectives
Future research into the effect of social jet lag should focus on a specific group of working people who have the highest exposure to social jet lag. Additionally, measurements of social jet lag should be performed at different points in time, which will allow a better insight into the long-term social jet lag that people experience, combined with a longer follow-up period for glycemic and metabolic changes to assess longitudinal effects. Elucidating the exact factors that are involved could offer new targets for prevention of metabolic disease.
In conclusion, our results show that social jet lag is cross-sectionally associated with the glycemic and metabolic markers HbA1c, fasting glucose, cholesterol ratio, systolic blood pressure, and BMI. Significant interaction with working status was present, and direction of the associations was generally detrimental in the working population, whereas higher social jet lag was associated with improved glycemic and metabolic control for retired people. Future studies should focus on the longitudinal effects of social jet lag on glycemic and metabolic control and explore the mechanisms and pathways that are involved.
ACKNOWLEDGMENTS
This study was made possible by collaboration with the Diabetes Care System West Friesland. The authors thank participants of this study and research staff of the Diabetes Care System West Friesland.
FUNDING INFORMATION
This study was supported by a Senior Fellowship grant from the Dutch Diabetes Foundation to Femke Rutters, grant number 2019.82.002.
CONFLICT OF INTEREST
The authors declared no conflict of interest.
Open Research
DATA AVAILABILITY STATEMENT
The data, code, and other materials that underlie the results reported in this article are available from [email protected] upon reasonable request to the Hoorn Steering Committee.
