Abstract
Background:
Evening chronotype may promote adherence to an unhealthy lifestyle and increase type 2 diabetes risk.
Objective:
To evaluate the role of modifiable lifestyle behaviors in the association between chronotype and diabetes risk.
Design:
Prospective cohort study.
Setting:
Nurses’ Health Study II.
Participants:
63 676 nurses aged 45 to 62 years with no history of cancer, cardiovascular disease, or diabetes in 2009 were prospectively followed until 2017.
Measurements:
Self-reported chronotype using a validated question from the Morningness-Eveningness Questionnaire. The lifestyle behaviors that were measured were diet quality, physical activity, alcohol intake, body mass index (BMI), smoking, and sleep duration. Incident diabetes cases were self-reported and confirmed using a supplementary questionnaire.
Results:
Participants reporting a “definite evening” chronotype were 54% (95% CI, 49% to 59%) more likely to have an unhealthy lifestyle than participants reporting a “definite morning” chronotype. A total of 1925 diabetes cases were documented over 469 120 person-years of follow-up. Compared with the “definite morning” chronotype, the adjusted hazard ratio (HR) for diabetes was 1.21 (CI, 1.09 to 1.35) for the “intermediate” chronotype and 1.72 (CI, 1.50 to 1.98) for the “definite evening” chronotype after adjustment for sociodemographic factors, shift work, and family history of diabetes. Further adjustment for BMI, physical activity, and diet quality attenuated the association comparing the “definite evening” and “definite morning” chronotypes to 1.31 (CI, 1.13 to 1.50), 1.54 (CI, 1.34 to 1.77), and 1.59 (CI, 1.38 to 1.83), respectively. Accounting for all measured lifestyle and sociodemographic factors resulted in a reduced but still positive association (HR comparing “definite evening” vs. “definite morning” chronotype, 1.19 [CI, 1.03 to 1.37]).
Limitations:
Chronotype assessment using a single question, self-reported data, and homogeneity of the study population.
Conclusion:
Middle-aged nurses with an evening chronotype were more likely to report unhealthy lifestyle behaviors and had increased diabetes risk compared with those with a morning chronotype. Accounting for BMI, physical activity, diet, and other modifiable lifestyle factors attenuated much but not all of the increased diabetes risk.
Primary Funding Source:
National Institutes of Health.
Chronotype, also known as circadian preference, is a partly genetically determined construct and refers to one’s inclination for earlier or later sleeping times (1–3). People who prefer to go to bed later at night (for example, 3:00 a.m.), get up later in the day (for example, 12:00 p.m.), or feel energetic later in the day (for example, 10:00 p.m.) are considered to have an evening chronotype (4). An estimated 8% of the population has an evening chronotype (5, 6), which has been linked to poor metabolic regulation, disrupted glycemic control, metabolic disorders, and higher incidence and prevalence of type 2 diabetes (7–9). However, the reasons for the observed association between evening chronotype and increased diabetes risk are not well understood.
Evening chronotype may increase diabetes risk through circadian misalignment, which occurs when circadian biological rhythms, including sleep–wake cycles, hormone secretion, body temperature regulation, and metabolism, are not synchronized with the physical (for example, light) and social (for example, work time) environment (10). Common situations for circadian misalignment include social jetlag (differences in sleep timing between weekdays and weekends), irregular or late sleep schedules, or night shift work (10). Circadian misalignment is particularly common in people with an evening chronotype (11) and may disrupt circadian rhythms in energy balance and metabolic regulation and promote diabetes onset (12, 13). In our prior study, increased diabetes risk associated with evening chronotype was highest among day workers, whereas the reduced diabetes risk associated with morning chronotype was not observed among those who worked night shift work for 10 or more years. This suggests that mismatch between chronotype and work timing may exacerbate the effect of chronotype on diabetes risk (14). Furthermore, genetic variants previously identified for chronotype are implicated in insulin secretion (for example, MADD) and obesity (for example, FTO) that could lead to diabetes development (6). Lifestyle behaviors may also play an important role in the association between chronotype and diabetes risk. Compared with morning chronotype, evening chronotype is associated with shorter and more irregular sleep (15, 16) and may negatively influence behaviors during the day, such as the amount and timing of eating and exercise (17, 18). There is also some evidence that people with an evening chronotype are more likely to engage in risky behaviors (such as smoking [19] and heavy drinking [20]) resulting from disrupted reward-related brain functions (11). However, the extent to which metabolic differences between morning and evening chronotypes are explained by their behavioral differences remains unclear (21).
Based on this conceptual framework (Supplement Figure 1), we aimed to investigate the role of modifiable lifestyle factors in the relationship between chronotype and diabetes risk. Specifically, we aimed to examine the associations between chronotype and lifestyle factors among middle-aged U.S. women and to assess changes in the relationship between evening chronotype and diabetes risk after accounting for modifiable lifestyle behaviors.
Methods
Study Design, Setting, and Participants
The Nurses’ Health Study II (NHSII) is an ongoing prospective cohort study initiated in 1989 among 116 429 female and predominantly White (>90%) registered nurses aged 25 to 42 years. Comprehensive questionnaires about lifestyle and health-related information were mailed to participants at baseline and every 2 years. The follow-up completion rate has been high (>85%) for each data collection cycle (22). We included 64 590 participants who answered the chronotype question in 2009, had information on lifestyle behaviors, and had no clinical diagnoses of diabetes, cancer, and cardiovascular disease in 2009. We further excluded participants who were lost to follow-up (n = 762) or had missing information on the date of diabetes diagnosis (n = 152), leaving 63 676 for analysis (Appendix Figure). The study protocol was approved by the Institutional Review Boards of Brigham and Women’s Hospital and Harvard T.H. Chan School of Public Health.
Chronotype
Chronotype was measured in 2009 and 2015 using 1 validated question (23) from the Morningness-Eveningness Questionnaire (4): “One hears about morning and evening types of people. Which ONE of these types do you consider yourself to be?” The response options were “Definitely a morning type”, “More of a morning than an evening type”, “More of an evening than a morning type”, “Definitely an evening type”, and “Neither”. This single question has been used in other large cohort studies (such as the UK Biobank [6]) to assess chronotype and has a correlation of 0.72 with the full score of the Morningness-Eveningness Questionnaire (24). Respondents and nonrespondents to the chronotype question shared similar characteristics, with more non-White participants, higher prevalent diabetes, and less healthy lifestyle behaviors among nonrespondents (Supplement Table 1). To reduce misclassification, we grouped the “neither”, “more morning”, and “more evening” types into an “intermediate” category, similar to what has been done in previous studies (14), resulting in 3 chronotype categories: definite morning, intermediate, and definite evening. Self-reported chronotype was found to have remained relatively stable over 6 years when the chronotype question was asked again in 2015 (the weighted κ coefficient showed moderate consistency of 0.626 when chronotype was grouped into 3 categories).
Lifestyle Behaviors
Six lifestyle behaviors assessed around 2009 were considered as outcomes in the analysis of associations between chronotype and lifestyle behaviors and as covariates in the analysis of prospective associations between chronotype and diabetes risk. Additional details about lifestyle assessment are provided in Supplement Table 2.
Diet
Diet was assessed via a 152-item semiquantitative food-frequency questionnaire asking how often participants had consumed specified amounts of different foods on average during the previous year (25). To measure overall diet quality, we used the Alternative Healthy Eating Index-2010 (AHEI) score without the alcohol component (26) and dichotomized this score at the top 40% of the distribution, as in previous studies (27–29). A score in the top 40% (≥61.8, based on the cohort-specific distribution) was defined as having a healthy diet.
Alcohol Use
Information on alcohol use was also collected with the food-frequency questionnaire. Low-risk alcohol use was defined as no to moderate drinking (<15 g/d, equivalent to up to 1 standard drink per day) (30).
Body Mass Index
Participants with a body mass index (BMI) of at least 18.5 kg/m2 but less than 25.0 kg/m2 were categorized as having a healthy body weight (31).
Physical Activity
Physical activity in the previous year was measured using a validated questionnaire and quantified as the total metabolic equivalent of task (MET) hours per week over 11 common recreational activities (32). The World Health Organization recommends at least 2.5 hours of moderate physical activity or at least 1.25 hours of vigorous physical activity per week, or an equivalent combination for adults (33). Assuming a MET of 3 for moderate activity and 6 for vigorous activity, we considered participants with a physical activity level of 7.5 MET-hours per week or higher to be physically active.
Smoking Status
Smoking status and history were self-reported. Low-risk smoking was defined as not currently smoking (29).
Sleep Duration
Participants self-reported their average number of hours of sleep in 24 hours. Those who reported sleeping at least 7 and less than 9 hours per day were placed in the healthy sleep duration category (34).
Healthy Lifestyle Score
Healthy Lifestyle Score (HLS) was calculated based on these 6 dichotomized factors; scores ranged from 0 to 6 (ordinal numbers), with 6 being the healthiest (28, 29). Following previous literature (27), participants with an HLS below 4 (median) were considered to have an overall unhealthy lifestyle.
Incident Diabetes
In biennial questionnaires, participants were asked whether they had received a clinical diagnosis of diabetes since the return of the last questionnaire. Participants who self-reported diabetes diagnoses were provided with a supplementary questionnaire about diagnostic tests, diabetes symptoms, and any hypoglycemic therapy. Participants who met 1 or more of the following 4 criteria were considered to have diabetes: 1) a diabetes symptom and either a fasting plasma glucose level of 7 mmol/L (126 mg/dL) or higher or a random plasma glucose level of 11.1 mmol/L (200 mg/dL) or higher, 2) 2 elevated plasma glucose levels defined by the same cutoffs on different occasions, 3) use of hypoglycemic medication, or 4) hemoglobin A1c level of 6.5% or higher for cases reported after 2010, in accordance with the updated American Diabetes Association guidelines (35–37). In a validation study of 62 randomly selected NHS participants who self-reported diabetes and completed the supplementary questionnaire, an endocrinologist reviewed their medical records and confirmed the diagnosis in 61 of the 62 participants (37).
Covariates
Birth date and race were self-reported in 1989. Diabetes diagnoses in first-degree relatives were assessed in 2009. Workplace (nursing department) was self-reported in 2009, 2011, and 2013. On every biennial questionnaire, information on menopausal status, use of postmenopausal hormone therapy, gestational diabetes (until 2001), aspirin use, and multivitamin use was collected, and residential address was updated and geocoded to derive geographic regions, population density, and census tract household income.
In 1989, participants self-reported the number of years they had worked rotating night shifts in their lifetime. On subsequent biennial questionnaires, participants self-reported the number of months in the previous 2 years they had worked rotating night shifts. We used these data to calculate lifetime duration of rotating night shifts (in months) (38) and any rotating night shift work in the previous 2 years in 2009, 2011, and 2013. In 2009, participants reported the average number of night shifts per month, and we imputed this information for subsequent cycles based on their shift work status.
Statistical Analysis
We estimated prevalence ratios (PRs) for having individual dichotomized unhealthy lifestyle behaviors or having an HLS less than 4 according to chronotype using Poisson regression with robust error variance (39, 40). In model 1, we adjusted for all of the aforementioned confounders except covariates related to shift work. Because rotating night shift work is an important source of circadian disruption in nurses, we further added any rotating night shift work in the previous 2 years, the average number of night shifts per month in the previous 2 years, and the cumulative months of rotating night shift work in model 2. For each lifestyle behavior, we further included the other 5 lifestyle variables in model 2 (mutual adjustment) to evaluate whether the observed associations between chronotype and individual lifestyles were independent of other lifestyle behaviors. The P value for trend was calculated by modeling the ordinal chronotype responses as a continuous variable.
In prospective analyses, we used Cox proportional hazards models with time-varying variables to estimate the hazard ratio (HR) for incident diabetes by chronotype categories, with person-time at risk from the return date of the 2009 questionnaire until diabetes diagnosis, death, or the last date of follow-up (June 2017). We tested the proportional hazards assumption using a likelihood ratio test, comparing models with and without interaction terms between chronotype and time variables. We did not find evidence of violation (P for interaction = 0.47 [Supplement Figure 2]). Here, all variables except sleep duration, race, family history of diabetes, and gestational diabetes were modeled as time-dependent. We fitted an age-adjusted model and 2 multivariable models adjusted for the covariates described earlier. To evaluate the effect of lifestyle behaviors on the association, we further adjusted model 2 for all 6 lifestyle behaviors as well as each lifestyle behavior individually. In subgroup analysis, we explored whether night shift work variables or dichotomized HLS modified the association between chronotype and incident diabetes by including an interaction term between continuous chronotype (ranging from 1 to 3) and each of the aforementioned variables (multiplicative effect heterogeneity analysis).
We conducted several sensitivity analyses. First, because obesity has a genetic basis and may reflect other chronic conditions, we created a secondary HLS that excluded BMI (unhealthy HLS cutoff, <3). We evaluated the association of this secondary HLS with chronotype while accounting for BMI as a covariate. Second, given the controversial association between alcohol consumption and diabetes risk, particularly in women (41), we considered an alternative HLS without alcohol consumption (unhealthy HLS cutoff, <3). Third, we redefined unhealthy drinking behavior as nonmoderate alcohol intake (<5 or ≥15 g/d) or as binge drinking (≥4 drinks per day) and repeated relevant analyses. Fourth, we restricted our analysis to participants reporting a consistent chronotype from 2009 to 2015. Finally, we used an alternative categorization of chronotype (definitely or more morning, intermediate, definitely or more evening) and repeated the analyses. We used SAS, version 9.4 (SAS Institute), for data analysis.
Role of the Funding Source
The funding sources had no role in the study design; collection, analysis, or interpretation of the data; writing of the manuscript; or the decision to submit the manuscript for publication.
Results
At baseline in 2009, the mean age of the study sample was 54 years (SD, 4.6), and most participants were White (97%), postmenopausal (67%), and from the Midwest (32%), with an average annual family income of $83 379 (Table 1). About 35% of the participants reported a definite morning chronotype, compared with 11% who reported a definite evening chronotype. Participants with different chronotypes had similar distributions in age and race (Table 1). Compared with participants with a definite morning chronotype, those with a definite evening chronotype were less likely to live in the Northeast or to work in an outpatient setting but more likely to have ever worked night shifts, to have worked any rotating night shift work in the previous 2 years, and to have depression. Unhealthy lifestyle behaviors were more common among participants with a definite evening chronotype (Supplement Figure 3).
Table 1.
Characteristics of Participants at Baseline, by Chronotype, in the Nurses’ Health Study II (2009)
| Characteristic | Overall Sample (n = 63 676) | Definite Morning Chronotype (n = 22 380) | Intermediate Chronotype (n = 34 167) | Definite Evening Chronotype (n = 7129) |
|---|---|---|---|---|
| Mean age (SD), y | 54.3 (4.6) | 54.6 (4.6) | 54.2 (4.6) | 54.2 (4.7) |
| Non-White, % | 3.2 | 3.1 | 3.1 | 3.5 |
| Mean gross household income (SD), $ | 83 379 (32 322) | 84 242 (33 048) | 83 099 (31 951) | 82 012 (31 718) |
| Region, % | ||||
| Northeast | 31.6 | 33.9 | 30.9 | 28.0 |
| Midwest | 32.4 | 30.7 | 33.4 | 32.8 |
| South | 19.6 | 18.9 | 19.7 | 21.5 |
| West | 16.4 | 16.5 | 16.0 | 17.7 |
| Mean population density per square kilometer (SD) | 962 (3179) | 980 (3183) | 929 (3047) | 1059 (3736) |
| Postmenopausal, % | 67.2 | 69.1 | 65.8 | 67.3 |
| Postmenopausal hormone therapy use, % | ||||
| Current | 15.4 | 15.3 | 15.4 | 15.4 |
| Past | 20.1 | 21.4 | 19.6 | 19.0 |
| Never | 64.5 | 63.3 | 65.0 | 65.6 |
| Family history of diabetes, % | 36.3 | 35.8 | 36.5 | 36.8 |
| Aspirin use, % | 31.1 | 31.3 | 30.9 | 31.8 |
| Multivitamin use, % | 58.5 | 59.0 | 58.5 | 57.0 |
| Ever rotating night shift work, % | 71.4 | 70.6 | 71.2 | 74.8 |
| Mean cumulative rotating night shift work (SD), mo* | 58.2 (55.1) | 56.0 (52.7) | 57.6 (54.6) | 67.5 (62.9) |
| Any rotating night shift work in previous 2 y, % | 8.5 | 5.1 | 8.3 | 19.7 |
| Mean number of night shifts per month (SD)† | 11.0 (5.8) | 10.2 (6.0) | 10.6 (5.8) | 12.4 (5.4) |
| Workplace, % | ||||
| ICU/ED/OR | 9.7 | 9.4 | 9.5 | 12.2 |
| In hospital | 19.5 | 18.4 | 19.8 | 21.6 |
| Outpatient | 31.0 | 32.4 | 31.5 | 24.4 |
| Retired/other | 39.8 | 39.8 | 39.3 | 41.9 |
| History of gestational diabetes, % | 3.9 | 3.5 | 4.0 | 4.5 |
| Clinical depression, % | 14.7 | 10.7 | 15.8 | 21.8 |
| Hypertension, % | 24.8 | 23.6 | 24.8 | 28.5 |
| Hypercholesterolemia, % | 33.2 | 31.3 | 33.6 | 37.2 |
| Lifestyle factors | ||||
| Mean Alternative Healthy Eating Index-2010 score (SD) | 57.4 (11.6) | 58.7 (11.5) | 57.0 (11.5) | 55.0 (11.9) |
| Mean BMI (SD), kg/m2 | 27.2 (6.0) | 26.5 (5.6) | 27.3 (6.1) | 28.6 (6.6) |
| Mean physical activity (SD), MET-h/wk | 24.4 (29.9) | 27.9 (33.8) | 23.1 (27.4) | 19.8 (26.7) |
| Current smoking, % | 5.8 | 5.2 | 5.6 | 9.0 |
| Mean alcohol use (SD), g/d | 6.5 (9.9) | 6.8 (9.9) | 6.5 (9.8) | 5.8 (9.9) |
| Sleep duration <7 or ≥9 h/d, % | 31.9 | 29.3 | 31.4 | 42.5 |
| Mean Healthy Lifestyle Score (SD)‡ | 4 (1.2) | 4.1 (1.1) | 3.9 (1.2) | 3.6 (1.2) |
BMI = body mass index; ED = emergency department; ICU = intensive care unit; MET = metabolic equivalent of task; OR = operating room.
Among participants who had ever worked rotating night shifts in 2009.
Among those who reported any rotating night shift work in the previous 2 years in 2009.
Healthy Lifestyle Score was calculated based on the 6 lifestyle factors and ranged from 0 to 6 (ordinal numbers), with 6 being the healthiest.
Associations Between Chronotype and Unhealthy Lifestyle
In fully adjusted models, associations were observed between chronotype and individual lifestyle behaviors, with the strongest magnitudes for current smoking (adjusted PR comparing definite evening vs. definite morning chronotype, 1.40 [95% CI, 1.27 to 1.54]), unhealthy sleep duration (adjusted PR, 1.29 [CI, 1.25 to 1.34]), and inadequate physical activity (adjusted PR, 1.21 [CI, 1.17 to 1.25]) (Table 2). Unhealthy BMI and low diet quality showed weaker but significant associations. We did not find a positive association between evening chronotype and unhealthy alcohol use (defined as ≥15 g/d). Compared with the definite morning chronotype, the PR for an overall unhealthy lifestyle (HLS <4) was 1.18 (CI, 1.16 to 1.21) for the intermediate chronotype and 1.54 (CI, 1.49 to 1.59) for the definite evening chronotype (P for trend < 0.001). We found weaker associations when we used the alternative chronotype categorization (Supplement Table 3). Using alternative definitions of HLS that excluded BMI or alcohol use did not alter the findings (Supplement Table 4). However, we observed modest positive associations between evening chronotype and unhealthy alcohol use when we defined it as nonmoderate consumption or binge drinking (Supplement Table 4).
Table 2.
Cross-Sectional Associations Between Chronotype and Unhealthy Lifestyle Factors, With Definite Morning Chronotype as the Reference Group (n = 63 676)
| Characteristic | Unhealthy Lifestyle/Sample Size, n/N | Prevalence Ratio (95% CI) |
||
|---|---|---|---|---|
| Model 1* | Model 2† | Model 2 + Mutual Adjustment‡ | ||
| Current smoking | ||||
| Definite morning chronotype | 1157/22 380 | 1.00 (reference) | 1.00 (reference) | 1.00 (reference) |
| Intermediate chronotype | 1924/34 167 | 1.09 (1.02–1.17)§ | 1.08 (1.01–1.16)§ | 1.02 (0.95–1.10) |
| Definite evening chronotype | 644/7129 | 1.70 (1.55–1.86)§ | 1.59 (1.45–1.75)§ | 1.40 (1.27–1.54)§ |
| P value for trend|| | – | <0.001§ | <0.001§ | <0.001§ |
| Sleep duration <7 or ≥9 h/d | ||||
| Definite morning chronotype | 6551/22 380 | 1.00 (reference) | 1.00 (reference) | 1.00 (reference) |
| Intermediate chronotype | 10 734/34 167 | 1.07 (1.04–1.10)§ | 1.06 (1.03–1.08)§ | 1.04 (1.02–1.07)§ |
| Definite evening chronotype | 3031/7129 | 1.42 (1.38–1.47)§ | 1.34 (1.29–1.38)§ | 1.29 (1.25–1.34)§ |
| P value for trend|| | – | <0.001§ | <0.001§ | <0.001§ |
| Physical activity <7.5 MET-h/wk | ||||
| Definite morning chronotype | 5915/22 380 | 1.00 (reference) | 1.00 (reference) | 1.00 (reference) |
| Intermediate chronotype | 10 707/34 167 | 1.19 (1.16–1.22)§ | 1.19 (1.16–1.22)§ | 1.10 (1.07–1.13)§ |
| Definite evening chronotype | 2800/7129 | 1.47 (1.42–1.53)§ | 1.46 (1.41–1.52)§ | 1.21 (1.17–1.25)§ |
| P value for trend|| | – | <0.001§ | <0.001§ | <0.001§ |
| BMI <18.5 or ≥25 kg/m2 | ||||
| Definite morning chronotype | 12 059/22 380 | 1.00 (reference) | 1.00 (reference) | 1.00 (reference) |
| Intermediate chronotype | 20 018/34 167 | 1.09 (1.07–1.10)§ | 1.08 (1.07–1.10)§ | 1.05 (1.03–1.06)§ |
| Definite evening chronotype | 4816/7129 | 1.24 (1.22–1.27)§ | 1.22 (1.20–1.25)§ | 1.14 (1.12–1.17)§ |
| P value for trend|| | – | <0.001§ | <0.001§ | <0.001§ |
| Low diet quality (AHEI score <61.8) | ||||
| Definite morning chronotype | 13 482/22 380 | 1.00 (reference) | 1.00 (reference) | 1.00 (reference) |
| Intermediate chronotype | 22 407/34 167 | 1.08 (1.06–1.09)§ | 1.07 (1.06–1.09)§ | 1.04 (1.03–1.06)§ |
| Definite evening chronotype | 5032/7129 | 1.16 (1.14–1.18)§ | 1.15 (1.13–1.17)§ | 1.08 (1.06–1.10)§ |
| P value for trend|| | – | <0.001§ | <0.001§ | <0.001§ |
| Alcohol intake ≥15 g/d | ||||
| Definite morning chronotype | 3014/22 380 | 1.00 (reference) | 1.00 (reference) | 1.00 (reference) |
| Intermediate chronotype | 4424/34 167 | 0.98 (0.94–1.03) | 0.99 (0.95–1.03) | 1.03 (0.99–1.07) |
| Definite evening chronotype | 811/7129 | 0.86 (0.80–0.93)§ | 0.90 (0.84–0.97)§ | 0.98 (0.91–1.06) |
| P value for trend|| | – | <0.001§ | 0.019§ | 0.77 |
| Unhealthy lifestyle (HLS <4) ¶ | ||||
| Definite morning chronotype | 6551/22 380 | 1.00 (reference) | 1.00 (reference) | NA |
| Intermediate chronotype | 11 975/34 167 | 1.19 (1.16–1.22)§ | 1.18 (1.16–1.21)§ | NA |
| Definite evening chronotype | 3394/7129 | 1.60 (1.55–1.65)§ | 1.54 (1.49–1.59)§ | NA |
| P value for trend|| | – | <0.001§ | <0.001§ | NA |
AHEI = Alternative Healthy Eating Index-2010; BMI = body mass index; HLS = Healthy Lifestyle Score; MET = metabolic equivalent of task; NA = not applicable.
Adjusted for age (months), race (white or non-White), gross household income (U.S. dollars), region (Northeast, Midwest, South, or West), quintile of population density (number of people per square kilometer), workplace (intensive care unit, emergency department, or operating room; in hospital; outpatient; or retired/other), family history of diabetes (yes or no), gestational diabetes (yes or no), aspirin use (yes or no), multivitamin use (yes or no), menopausal status (yes or no), and postmenopausal hormone use (current, past, or never).
Adjusted for covariates in model 1 plus cumulative duration of night shift work (months), any night shift work in the previous 2 years (yes or no), and average monthly number of night shifts.
Adjusted for covariates in model 2 plus the other 5 lifestyle behaviors.
P < 0.05.
Calculated by modeling chronotype as a continuous variable ranging from 1 to 3.
HLS was calculated based on the 6 lifestyle factors and ranged from 0 to 6 (ordinal numbers), with 6 being the healthiest. We dichotomized this score at the median (<4 vs. ≥4). Persons with an HLS <4 were considered to have an overall unhealthy lifestyle.
Association Between Chronotype and Incident Diabetes
In prospective analysis, 1925 incident diabetes cases were documented during 469 120 person-years of follow-up (mean, 7.4 years; Table 3). After adjustment for the covariates in model 2, the HR for diabetes risk compared with definite morning chronotype was 1.21 (CI, 1.09 to 1.35) for the intermediate chronotype and 1.72 (CI, 1.50 to 1.98) for the definite evening chronotype. This association was notably attenuated after additional adjustment for each of the 6 lifestyle behaviors. The strongest attenuation was observed for BMI (adjusted HR comparing the definite evening vs. the definite morning chronotype, 1.31 [CI, 1.13 to 1.50]), followed by physical activity (adjusted HR, 1.54 [CI, 1.34 to 1.77]), and the weakest attenuation was observed for smoking (adjusted HR, 1.71 [CI, 1.49 to 1.97]). Overall, with adjustment for all 6 lifestyle behaviors together, the association between chronotype and incident diabetes noticeably diminished, though it remained positive (adjusted HR, 1.19 [CI, 1.03 to 1.37]). We found weaker associations with diabetes risk when using the alternative chronotype categorization (Supplement Table 5). Associations were slightly stronger when we restricted the analysis to participants who were placed in identical chronotype categories in 2009 and 2015 (Supplement Table 6). The adjusted HR comparing the definite evening chronotype versus the definite morning chronotype was 2.17 (CI, 1.77 to 2.66) before adjustment for lifestyle behaviors and 1.32 (CI, 1.07 to 1.64) after adjustment for all 6 lifestyle behaviors.
Table 3.
Chronotype, Unhealthy Lifestyle, and Diabetes Risk in the Nurses’ Health Study II (n = 63 676)
| Variable | Definite Morning Chronotype | Intermediate Chronotype | Definite Evening Chronotype | Overall Sample |
|---|---|---|---|---|
| Cases, n/person-years | 557/163 054 | 1045/253 321 | 323/52 744 | 1925/469 120 |
| Crude incidence per 100 000 person-years | 342 | 413 | 612 | 410 |
| Model | Hazard Ratio (95% CI) |
P Value for Trend | ||
| Definite Morning Chronotype | Intermediate Chronotype | Definite Evening Chronotype | ||
| Age-adjusted model | 1.00 (reference) | 1.24 (1.12–1.37)* | 1.86 (1.62–2.14)* | <0.001* |
| Model 1† | 1.00 (reference) | 1.22 (1.10–1.35)* | 1.77 (1.54–2.03)* | <0.001* |
| Model 2‡ | 1.00 (reference) | 1.21 (1.09–1.35)* | 1.72 (1.50–1.98)* | <0.001* |
| Model 2 + body mass index | 1.00 (reference) | 1.06 (0.96–1.18) | 1.31 (1.13–1.50)* | <0.001* |
| Model 2 + physical activity§ | 1.00 (reference) | 1.14 (1.03–1.27)* | 1.54 (1.34–1.77)* | <0.001* |
| Model 2 + diet quality§ | 1.00 (reference) | 1.18 (1.06–1.30)* | 1.59 (1.38–1.83)* | <0.001* |
| Model 2 + alcohol use§ | 1.00 (reference) | 1.20 (1.08–1.33)* | 1.68 (1.46–1.93)* | <0.001* |
| Model 2 + sleep duration | 1.00 (reference) | 1.21 (1.09–1.34)* | 1.68 (1.46–1.93)* | <0.001* |
| Model 2 + smoking status | 1.00 (reference) | 1.21 (1.09–1.34)* | 1.71 (1.49–1.97)* | <0.001* |
| Model 2 + all lifestyle behaviors§ | 1.00 (reference) | 1.02 (0.92–1.14) | 1.19 (1.03–1.37)* | 0.035* |
P < 0.05.
Adjusted for age (months), race (white or non-White), gross household income (U.S. dollars), region (Northeast, Midwest, South, or West), quintile of population density (number of people per square kilometer), workplace (intensive care unit, emergency department, or operating room; in hospital; outpatient; or retired/other), family history of diabetes (yes or no), gestational diabetes (yes or no), aspirin use (yes or no), multivitamin use (yes or no), menopausal status (yes or no), and postmenopausal hormone use (current, past, or never).
Adjusted for covariates in model 1 plus cumulative duration of night shift work (months), any night shift work in the previous 2 years (yes or no), and average monthly number of night shifts.
To minimize residual confounding, we adjusted for continuous body mass index (weight in kilograms divided by square of height in meters), continuous physical activity (metabolic equivalent of task hours per week), continuous diet quality score (Alternative Healthy Eating Index-2010), and continuous alcohol intake (grams per day).
In subgroup analysis, although we did not observe statistically significant differences by shift work variables (P for interaction ≥ 0.149; Table 4), the association between chronotype and diabetes in the fully adjusted model was most apparent among participants with no night shift work in the previous 2 years or less than 10 years of lifetime night shift work. The association between chronotype and diabetes risk was similar by the HLS (P for interaction = 0.33).
Table 4.
Associations Between Chronotype and Diabetes Risk Among Subgroups of the Study Sample*
| Subgroup | Cases, n/Person-Years | Hazard Ratio (95% CI) |
P Value for Trend | P Value for Interaction | ||
|---|---|---|---|---|---|---|
| Definite Morning Chronotype | Intermediate Chronotype | Definite Evening Chronotype | ||||
| Any night shift work in previous 2 y | 0.149 | |||||
| Yes | 155/30 548 | 1.00 (reference) | 1.00 (0.62–1.61) | 0.91 (0.51–1.61) | 0.72 | |
| No | 1770/438 572 | 1.00 (reference) | 1.01 (0.91–1.13) | 1.23 (1.06–1.42)† | 0.024† | |
| Cumulative amount of rotating night shift work | 0.173 | |||||
| None | 457/133 197 | 1.00 (reference) | 0.85 (0.69–1.05) | 0.99 (0.72–1.36) | 0.52 | |
| <10 y | 1237/293 736 | 1.00 (reference) | 1.12 (0.98–1.27) | 1.32 (1.10–1.58)† | 0.003† | |
| ≥10 y | 231/42 187 | 1.00 (reference) | 0.86 (0.61–1.20) | 0.86 (0.56–1.32) | 0.45 | |
| Average number of night shifts per month | 0.25 | |||||
| No night shift work in previous 2 y | 1770/438 572 | 1.00 (reference) | 1.01 (0.91–1.13) | 1.23 (1.06–1.42)† | 0.024† | |
| <14 | 106/26 049 | 1.00 (reference) | 1.00 (0.56–1.79) | 0.94 (0.47–1.91) | 0.87 | |
| ≥14 | 49/4498 | 1.00 (reference) | 0.54 (0.06–4.72) | 1.75 (0.18–17.05) | 0.27 | |
| Healthy Lifestyle Score | 0.33 | |||||
| <4 | 1123/150 581 | 1.00 (reference) | 1.01 (0.88–1.17) | 1.18 (0.99–1.42) | 0.098 | |
| ≥4 | 802/318 539 | 1.00 (reference) | 1.03 (0.88–1.20) | 1.24 (0.97–1.57) | 0.145 | |
All models were adjusted for age (months), race (white or non-White), gross household income (U.S. dollars), region (Northeast, Midwest, South, or West), quintile of population density (number of people per square kilometer), workplace (intensive care unit, emergency department, or operating room; in hospital; outpatient; or retired/other), family history of diabetes (yes or no), gestational diabetes (yes or no), aspirin use (yes or no), multivitamin use (yes or no), menopausal status (yes or no), postmenopausal hormone use (current, past, or never), cumulative duration of night shift work (months), any night shift work in the previous 2 years (yes or no), average monthly number of night shifts, continuous diet quality score (Alternative Healthy Eating Index-2010), continuous body mass index (weight in kilograms divided by square of height in meters), continuous physical activity (metabolic equivalent of task hours per week), continuous alcohol intake (grams per day), dichotomized sleep duration, and dichotomized smoking status. If the stratification variable was continuous, we further adjusted for the continuous variable in the stratified models to reduce residual confounding.
P < 0.05.
Discussion
Compared with the definite morning chronotype, participants with the definite evening chronotype were more likely to have unhealthy lifestyle behaviors and develop diabetes over 7.4 years of follow-up independent of other risk factors, including rotating night shift work. Although evening chronotype was most strongly associated with smoking and short or long sleep duration, adjustment for BMI, physical activity, or diet quality led to the most notable attenuation of the prospective association between chronotype and diabetes risk. Full adjustment for all measured lifestyle and sociodemographic factors resulted in an attenuated but still positive association between evening chronotype and diabetes risk. In subgroup analysis, the fully adjusted association between evening chronotype and diabetes risk was stronger among women who had no night shift work in the previous 2 years or had a lifetime duration of night shift work of less than 10 years.
Similar to our findings, prior systematic reviews and observational studies have reported associations between evening chronotype and individual unhealthy lifestyle behaviors, such as low diet quality (42), frequent alcohol (20) and nicotine (19) use, increased BMI (43), physical inactivity (18), and sleep problems (15, 16). However, few if any studies have evaluated multiple lifestyle behaviors simultaneously. Notably, although evening chronotype was not associated with unhealthy drinking (defined as ≥1 drink per day), in secondary analyses, we found that evening chronotype was associated with unhealthy alcohol consumption (defined as binge drinking equivalent to ≥4 drinks per day). Taken together, our findings suggest that persons with the evening chronotype are more likely to be characterized by clustering of unhealthy lifestyle behaviors, which may help future public health programs identify targeted populations for lifestyle intervention.
A recent systematic review of 7 cross-sectional studies reported that evening chronotype is associated with a worse cardiometabolic profile and diabetes outcome (7). Our study adds important prospective evidence supporting a positive association between eveningness and diabetes risk, highlighting 72% increased diabetes incidence between persons with definite evening versus definite morning chronotype. Adjustment for lifestyle factors, particularly BMI, physical activity, and diet quality, substantially attenuated this relationship, suggesting a crucial role of lifestyle behaviors in the relationship between evening chronotype and diabetes risk. Interestingly, a previous study among patients with diabetes found partial mediation by higher caloric intake at dinner for the association between late chronotype and poor glycemic control (9), although this study did not assess other lifestyle behaviors or an overall lifestyle pattern. Given the importance of lifestyle modification in diabetes prevention, future research is warranted to investigate whether improving lifestyle behaviors could effectively reduce diabetes risk in persons with an evening chronotype.
Consistent with our prior study (14), our subgroup analysis only found the positive association between evening chronotype and diabetes risk in persons who had not worked night shifts recently but not in those who had. This could have mechanistic implications regarding circadian misalignment resulting from mismatch between chronotype (for example, late circadian preference) and work schedules (for example, daytime schedules) in diabetes development. When we examined lifetime duration of night shift work, the increased diabetes risk associated with evening chronotype was apparent only in those who had worked less than 10 years of night shifts in the past. Although changes in work schedules may alter chronotype, this observation may also be explained by potential reverse causation in that the underlying health reasons for quitting night shift work (such as weight gain) increased diabetes risk. Additional studies are needed to elucidate the genetic, neuroendocrine, and behavioral mechanisms through which chronotype interacts with work schedules to influence diabetes risk and the benefits of personalized shift scheduling tailored to the worker’s chronotype.
Strengths of our study include its large population-based sample, prospective cohort design, long-term follow-up, high retention rates, and repeated covariate assessment. Notably, simultaneous assessment of multiple lifestyle behaviors, which have been extensively validated in our cohort (25, 32, 44), allowed us to evaluate the association between chronotype and an overall lifestyle pattern that had not been addressed in previous studies.
Caution should be exercised when generalizing our findings. Our study population included primarily middle-aged White female nurses with relatively high education and socioeconomic status and a specific working environment. Future investigation in other populations is needed to determine whether our findings are applicable to men, non-White racial or ethnic groups, or other socioeconomic classes. Moreover, generational differences in diet, exercise, and body weight may limit the applicability of our findings to younger or older generations or current times.
Other limitations of the study included use of a single question to assess chronotype and reliance on self-reported data, possibly resulting in misclassification and measurement error. However, the correlation between this representative question and the overall Morningness-Eveningness Questionnaire score is high, ranging from 0.72 (24) to 0.88 (45). Although nondifferential misclassification in self-reported data might have biased the associations in either direction, more likely toward the null, the misclassification was likely minimal given our extensive validation of these self-reported data. For example, a prior validation study found diabetes prevalence below 1% among a random sample of NHS participants who did not self-report diabetes diagnoses (46). Finally, although the evening chronotype and the associated late sleep timing may be upstream drivers of unhealthy lifestyle behaviors, our cross-sectional analysis cannot fully preclude the possibility that lifestyle behaviors may influence how participants perceived and self-reported their chronotype.
In summary, middle-aged women with an evening chronotype were more likely to have unhealthy lifestyle behaviors and factors compared with those with a morning chronotype, which may contribute to their increased risk for type 2 diabetes. The prospective association between evening chronotype and diabetes risk became weaker after adjustment for unhealthy lifestyle behaviors, mainly BMI, physical activity, and diet quality. Accounting for all measured sociodemographic and lifestyle factors resulted in a reduced but still positive association between evening chronotype and diabetes risk, which was primarily observed among day workers. Future studies are needed to assess whether lifestyle interventions and personalized shift scheduling could reduce the adverse effect of evening chronotype on diabetes risk.
Supplementary Material
Acknowledgment:
The authors thank the women participating in the NHSII, who provided the information about their lives that enabled this study to be done.
Financial Support:
This study was supported by grants U01 CA176726 and R01HL155395 from the National Institutes of Health (NIH). Dr. Huang is supported by NIH grant K01HL143034. Dr. Schernhammer was supported by funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation program (grant agreement no. 101053225).
The study protocol was approved by the Institutional Review Boards of Brigham and Women’s Hospital and Harvard T.H. Chan School of Public Health. Participants’ consent was implied when they returned the questionnaires.
Footnotes
Disclaimer: The views and opinions expressed in this article are those of the authors and do not necessarily reflect those of the National Institutes of Health, the European Union, or the European Research Council Executive Agency.
Disclosures: Disclosures can be viewed at www.acponline.org/authors/icmje/ConflictOfInterestForms.do?msNum=M23-0728.
Reproducible Research Statement: Study protocol and data set: Available on request (https://nurseshealthstudy.org). Statistical code: Available from Dr. Kianersi (e-mail, nhkia@channing.harvard.edu; skianersi@bwh.harvard.edu).
Contributor Information
Sina Kianersi, Channing Division of Network Medicine, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, Massachusetts.
Yue Liu, Channing Division of Network Medicine, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, Massachusetts.
Marta Guasch-Ferré, Department of Nutrition, Harvard T.H. Chan School of Public Health, Boston, Massachusetts, and Department of Public Health and Novo Nordisk Foundation Center for Basic Metabolic Research, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark.
Susan Redline, Division of Sleep Medicine, Harvard Medical School; Division of Sleep and Circadian Disorders, Department of Medicine, Brigham and Women’s Hospital; and Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, Massachusetts.
Eva Schernhammer, Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, Massachusetts, and Department of Epidemiology, Center for Public Health, Vienna, Austria.
Qi Sun, Channing Division of Network Medicine, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, and Department of Epidemiology and Department of Nutrition, Harvard T.H. Chan School of Public Health, Boston, Massachusetts.
Tianyi Huang, Channing Division of Network Medicine, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, and Division of Sleep Medicine, Harvard Medical School, Boston, Massachusetts.
References
- 1.Roenneberg T. What is chronotype? Sleep and Biological Rhythms. 2012;10:75–76. [ 10.1111/j.1479-8425.2012.00541.x] [DOI] [Google Scholar]
- 2.Roenneberg T, Merrow M. Entrainment of the human circadian clock. Cold Spring Harb Symp Quant Biol. 2007;72:293–299. [PMC][ 10.1101/sqb.2007.72.043] [18419286] [DOI] [PubMed] [Google Scholar]
- 3.Koskenvuo M, Hublin C, Partinen M, et al. Heritability of diurnal type: a nationwide study of 8753 adult twin pairs. J Sleep Res. 2007;16:156–162. [PMC][ 10.1111/j.1365-2869.2007.00580.x] [17542945] [DOI] [PubMed] [Google Scholar]
- 4.Horne JA, Ostberg O. A self-assessment questionnaire to determine morningness-eveningness in human circadian rhythms. Int J Chronobiol. 1976;4:97–110. [1027738] [PubMed] [Google Scholar]
- 5.Fischer D, Lombardi DA, Marucci-Wellman H, et al. Chronotypes in the US - influence of age and sex. PLoS One. 2017;12:e0178782. [PMC][ 10.1371/journal.pone.0178782] [28636610] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Jones SE, Lane JM, Wood AR, et al. Genome-wide association analyses of chronotype in 697,828 individuals provides insights into circadian rhythms. Nat Commun. 2019;10:343. [PMC][ 10.1038/s41467-018-08259-7] [30696823] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Lotti S, Pagliai G, Colombini B, et al. Chronotype differences in energy intake, cardiometabolic risk parameters, cancer, and depression: a systematic review with meta-analysis of observational studies. Adv Nutr. 2022;13:269–281. [PMC][ 10.1093/advances/nmab115] [34549270] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Merikanto I, Lahti T, Puolijoki H, et al. Associations of chronotype and sleep with cardiovascular diseases and type 2 diabetes. Chronobiol Int. 2013;30:470–477. [PMC][ 10.3109/07420528.2012.741171] [23281716] [DOI] [PubMed] [Google Scholar]
- 9.Reutrakul S, Hood MM, Crowley SJ, et al. Chronotype is independently associated with glycemic control in type 2 diabetes. Diabetes Care. 2013;36:2523–2529. [PMC][ 10.2337/dc12-2697] [23637357] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Roenneberg T, Merrow M. The circadian clock and human health. Curr Biol. 2016;26:R432–R443. [PMC][ 10.1016/j.cub.2016.04.011] [27218855] [DOI] [PubMed] [Google Scholar]
- 11.Baron KG, Reid KJ. Circadian misalignment and health. Int Rev Psychiatry. 2014;26:139–154. [PMC][ 10.3109/09540261.2014.911149] [24892891] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Poggiogalle E, Jamshed H, Peterson CM. Circadian regulation of glucose, lipid, and energy metabolism in humans. Metabolism. 2018;84:11–27. [PMC][ 10.1016/j.metabol.2017.11.017] [29195759] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Cribbet MR, Logan RW, Edwards MD, et al. Circadian rhythms and metabolism: from the brain to the gut and back again. Ann N Y Acad Sci. 2016;1385:21–40. [PMC][ 10.1111/nyas.13188] [27589593] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Vetter C, Devore EE, Ramin CA, et al. Mismatch of sleep and work timing and risk of type 2 diabetes. Diabetes Care. 2015;38:1707–1713. [PMC][ 10.2337/dc15-0302] [26109502] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Baden MY, Hu FB, Vetter C, et al. Sleep duration patterns in early to middle adulthood and subsequent risk of type 2 diabetes in women. Diabetes Care. 2020;43:1219–1226. [PMC][ 10.2337/dc19-2371] [32209646] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Huang T, Redline S. Cross-sectional and prospective associations of actigraphy-assessed sleep regularity with metabolic abnormalities: the Multi-Ethnic Study of Atherosclerosis. Diabetes Care. 2019;42:1422–1429. [PMC][ 10.2337/dc19-0596] [31167888] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Mazri FH, Manaf ZA, Shahar S, et al. The association between chronotype and dietary pattern among adults: a scoping review. Int J Environ Res Public Health. 2019;17:68. [PMC][ 10.3390/ijerph17010068] [31861810] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Mota MC, Waterhouse J, De-Souza DA, et al. Association between chronotype, food intake and physical activity in medical residents. Chronobiol Int. 2016;33:730–739. [PMC][ 10.3109/07420528.2016.1167711] [27096153] [DOI] [PubMed] [Google Scholar]
- 19.Wittmann M, Paulus M, Roenneberg T. Decreased psychological well-being in late ‘chronotypes’ is mediated by smoking and alcohol consumption. Subst Use Misuse. 2010;45:15–30. [PMC][ 10.3109/10826080903498952] [20025436] [DOI] [PubMed] [Google Scholar]
- 20.Chronotype Adan A. and personality factors in the daily consumption of alcohol and psychostimulants. Addiction. 1994;89:455–462. [PMC][ 10.1111/j.1360-0443.1994.tb00926.x] [8025504] [DOI] [PubMed] [Google Scholar]
- 21.Stenvers DJ, Scheer F, Schrauwen P, et al. Circadian clocks and insulin resistance. Nat Rev Endocrinol. 2019;15:75–89. [PMC][ 10.1038/s41574-018-0122-1] [30531917] [DOI] [PubMed] [Google Scholar]
- 22.Bao Y, Bertoia ML, Lenart EB, et al. Origin, methods, and evolution of the three Nurses’ Health Studies. Am J Public Health. 2016;106:1573–1581. [PMC][ 10.2105/AJPH.2016.303338] [27459450] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Zavada A, Gordijn MC, Beersma DG, et al. Comparison of the Munich Chronotype Questionnaire with the Horne-Ostberg’s Morningness-Eveningness Score. Chronobiol Int. 2005;22:267–278. [PMC][ 10.1081/cbi-200053536] [16021843] [DOI] [PubMed] [Google Scholar]
- 24.Megdal SP, Schernhammer ES. Correlates for poor sleepers in a Los Angeles high school. Sleep Med. 2007;9:60–63. [PMC][ 10.1016/j.sleep.2007.01.012] [17869576] [DOI] [PubMed] [Google Scholar]
- 25.Yuan C, Spiegelman D, Rimm EB, et al. Validity of a dietary questionnaire assessed by comparison with multiple weighed dietary records or 24-hour recalls. Am J Epidemiol. 2017;185:570–584. [PMC][ 10.1093/aje/kww104] [28338828] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Chiuve SE, Fung TT, Rimm EB, et al. Alternative dietary indices both strongly predict risk of chronic disease. J Nutr. 2012;142:1009–1018. [PMC][ 10.3945/jn.111.157222] [22513989] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Li Y, Schoufour J, Wang DD, et al. Healthy lifestyle and life expectancy free of cancer, cardiovascular disease, and type 2 diabetes: prospective cohort study. BMJ. 2020;368:l6669. [PMC][ 10.1136/bmj.l6669] [31915124] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Li Y, Pan A, Wang DD, et al. Impact of healthy lifestyle factors on life expectancies in the US population. Circulation. 2018;138:345–355. [PMC][ 10.1161/CIRCULATIONAHA.117.032047] [29712712] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Guasch-Ferre M, Li Y, Bhupathiraju SN, et al. Healthy lifestyle score including sleep duration and cardiovascular disease risk. Am J Prev Med. 2022;63:33–42. [PMC][ 10.1016/j.amepre.2022.01.027] [35361505] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.U.S. Department of Health and Human Services; U.S. Department of Agriculture. 2015–2020 Dietary Guidelines for Americans. 8th edition. December 2015. Accessed at https://health.gov/our-work/food-nutrition/previous-dietary-guidelines/2015 on 29 May 2023. [Google Scholar]
- 31.Centers for Disease Control and Prevention. Defining Adult Overweight & Obesity. Accessed at www.cdc.gov/obesity/basics/adult-defining.html on 11 July 2022. [Google Scholar]
- 32.Wolf AM, Hunter DJ, Colditz GA, et al. Reproducibility and validity of a self-administered physical activity questionnaire. Int J Epidemiol. 1994;23:991–999. [PMC][ 10.1093/ije/23.5.991] [7860180] [DOI] [PubMed] [Google Scholar]
- 33.Bull FC, Al-Ansari SS, Biddle S, et al. World Health Organization 2020 guidelines on physical activity and sedentary behaviour. Br J Sports Med. 2020;54:1451–1462. doi: 10.1136/bjsports-2020-102955 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Lloyd-Jones DM, Allen NB, Erson CAM, et al. ; American Heart Association. Life’s Essential 8: updating and enhancing the American Heart Association’s construct of cardiovascular health: a presidential advisory from the American Heart Association. Circulation. 2022;146:e18–e43. [PMC][ 10.1161/CIR.0000000000001078] [35766027] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.National Diabetes Data Group. Classification and diagnosis of diabetes mellitus and other categories of glucose intolerance. Diabetes. 1979;28:1039–1057. [PMC][ 10.2337/diab.28.12.1039] [510803] [DOI] [PubMed] [Google Scholar]
- 36.American Diabetes Association. Diagnosis and classification of diabetes mellitus. Diabetes Care. 2014;37 Suppl 1:S81–S90. [PMC][ 10.2337/dc14-S081] [24357215] [DOI] [PubMed] [Google Scholar]
- 37.Manson JE, Rimm EB, Stampfer MJ, et al. Physical activity and incidence of non-insulin-dependent diabetes mellitus in women. Lancet. 1991;338:774–778. [PMC][ 10.1016/0140-6736(91)90664-b] [1681160] [DOI] [PubMed] [Google Scholar]
- 38.Anothaisintawee T, Reutrakul S, Van Cauter E, et al. Sleep disturbances compared to traditional risk factors for diabetes development: systematic review and meta-analysis. Sleep Med Rev. 2016;30:11–24. [PMC][ 10.1016/j.smrv.2015.10.002] [26687279] [DOI] [PubMed] [Google Scholar]
- 39.Zou G. A modified Poisson regression approach to prospective studies with binary data. Am J Epidemiol. 2004;159:702–706. [PMC][ 10.1093/aje/kwh090] [15033648] [DOI] [PubMed] [Google Scholar]
- 40.Barros AJ, Hirakata VN. Alternatives for logistic regression in cross-sectional studies: an empirical comparison of models that directly estimate the prevalence ratio. BMC Med Res Methodol. 2003;3:21. [PMC][ 10.1186/1471-2288-3-21] [14567763] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Neuenschwander M, Ballon A, Weber KS, et al. Role of diet in type 2 diabetes incidence: umbrella review of meta-analyses of prospective observational studies. BMJ. 2019;366:l2368. [PMC][ 10.1136/bmj.l2368] [31270064] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Teixeira GP, Guimarães KC, Soares AGN, et al. Role of chronotype in dietary intake, meal timing, and obesity: a systematic review. Nutr Rev. 2022;81:75–90. [PMC][ 10.1093/nutrit/nuac044] [35771674] [DOI] [PubMed] [Google Scholar]
- 43.van der Merwe C, Münch M, Kruger R. Chronotype differences in body composition, dietary intake and eating behavior outcomes: a scoping systematic review. Adv Nutr. 2022;13:2357–2405. [PMC][ 10.1093/advances/nmac093] [36041181] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Al-Delaimy WK, Willett WC. Measurement of tobacco smoke exposure: comparison of toenail nicotine biomarkers and self-reports. Cancer Epidemiol Biomarkers Prev. 2008;17:1255–1261. [PMC][ 10.1158/1055-9965.EPI-07-2695] [18483348] [DOI] [PubMed] [Google Scholar]
- 45.Maukonen M, Havulinna AS, Mannisto S, et al. Genetic associations of chronotype in the Finnish general population. J Biol Rhythms. 2020;35:501–511. [PMC][ 10.1177/0748730420935328] [32579418] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Field AE, Coakley EH, Must A, et al. Impact of overweight on the risk of developing common chronic diseases during a 10-year period. Arch Intern Med. 2001;161:1581–1586. [PMC][ 10.1001/archinte.161.13.1581] [11434789] [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.