Misalignment between circadian preference and accelerometer-derived actual sleep-wake cycle is associated with increased risk of cardiometabolic diseases: a prospective cohort study in UK Biobank =================================================================================================================================================================================================== * Yue Chen * Tingting Geng * Xinming Xu * Zhicheng Zhang * Lili Huang * Haiyang Dong * Huan Yu * Xiang Gao * Liang Sun ## Abstract **Background** Evidence has linked several circadian disruption indicators, such as social jetlag and shift work, to adverse health outcomes, however, associations of misalignment between circadian preference and actual sleep-wake cycle with cardiometabolic diseases (CMDs) remains unclear. We aimed to prospectively investigate the associations of the circadian misalignment with CMDs including type 2 diabetes (T2D), chronic heart diseases (CHD), and stroke, and to uncover potential mechanisms linking circadian misalignment and CMDs. **Methods** A total of 60,965 participants from the UK Biobank study without baseline CMDs and followed-up for an average of 7.9 years were included in the current analysis. Circadian misalignment was defined as discrepancies between self-reported chronotype and accelerometer-derived midpoint of sleep to detect its association with CMDs. Incident CMDs were derived from primary care, hospital inpatient, death registry, and self-reported source of data. Cox proportional hazards regression models were used to compute the hazard ratio (HR) and confidence intervals (CIs) on the association between circadian misalignment (quintiles of residuals of midsleep on chronotype) and incident CMDs. **Results** U-shaped associations were found of the circadian misalignment with incident T2D and CHD after adjusting the potential confounders. Compared to individuals with aligned midsleep and circadian preferences (Q3), those with advanced and delayed circadian misalignment had higher risks of T2D [HR (95%CI) 1.22 (1.03, 1.45) in Q1 and 1.39 (1.18, 1.62) in Q5]. However, only delayed circadian misalignment was significantly associated with an increased risk of CHD [HR (95%CI) 1.15 (1.01, 1.31) in Q4 and 1.16 (1.02, 1.33) in Q5]. Liver function, lipid and glucose metabolism, and inflammatory markers partially explained the observed circadian misalignment and CMDs association (mediation proportion 12.3-44.6% for T2D, 8.8-20.5% for CHD). Moreover, the association between delayed circadian misalignment and CMDs was more prominent in women (for T2D, *Pinteraction*=0.03) and in younger adults (for CHD, *Pinteraction*=0.02) compared to their counterparts. Additionally, early chronotype [HR (95%CI): 1.19 (1.06, 1.34)] rather than late chronotype was associated with an increased risk of incident T2D. **Conclusion** Both advanced and delayed circadian misalignment were associated with increased risks of CMDs, suggesting potential benefits of aligning actual sleep-wake cycles with individual circadian preferences. **What is known?** * Cardiometabolic diseases such as type 2 diabetes, coronary heart disease, and cerebrovascular diseases have become major concerns in global public health over recent decades. * Evidence has linked several circadian disruption indicators to adverse health outcomes: People living with late chronotype (circadian preference), late sleep timing, long social jetlag (the difference between midsleep on work-free days and work days), and shift work were prone to introduce circadian disruptions and had higher risks of cardiometabolic diseases. * Little is known about whether and how the misalignment between circadian preference and actual sleep-wake cycle is associated with cardiometabolic diseases, particularly in large prospective cohort studies. **What New Information Does This Article Contribute?** * This is the first study to propose a new metrics (misalignment between circadian preference and actual sleep-wake cycle) to shed some lights on quantifying human circadian disruption in general population of differed work schedule and in population with less disrupted circadian rhythm. * Both advanced and delayed sleep-wake cycle comparing to individual circadian preference were associated with increased risks of cardiometabolic diseases. Population may benefit from keeping an actual sleep-wake cycle in accordance with their circadian preference. * Early chronotype rather than late chronotype were found to associate with an increased risk of incident T2D, independent of actual sleep-wake cycle. * Age and sex modified the associations between circadian misalignment and cardiometabolic diseases, while liver function, lipid and glucose metabolism, and inflammatory markers partially mediated the associations. Keywords * circadian misalignment * chronotype * diabetes * cardiometabolic diseases * UK Biobank ## Introduction Cardiometabolic diseases (CMDs) such as type 2 diabetes (T2D), coronary heart disease (CHD), and cerebrovascular diseases have become major concerns in global public health over recent decades1. The multifactorial nature of the etiology of these diseases involves a complex interplay of genetic, lifestyle, and environmental factors, presenting great challenges on the prevention and prognosis of CMDs2. Therefore, more efforts are needed to develop potential strategies to combat this escalating phenomenon. Circadian disruption indicated a wide range of misalignment upon individual internal-internal timing as well as internal-external rhythm3. In field-based studies, metrics for circadian disruption mainly rely on rest-activity measurement and disrupted sleep behavior, e.g., social jetlag and shift work3. Epidemiological evidence suggested that people living with late chronotype (circadian preference), late sleep timing, long social jetlag (the difference between midsleep on work-free days and work days), and shift work were prone to introduce circadian disruptions4 and had higher risks of CMDs5-9. However, current widely used proxies for circadian disruption (such as social jetlag and shift work) either largely require regular work schedule, or lack of generalizability to the general population who works non-shift and less deviated from internal clock10. In addition, the majority of prior epidemiological studies on circadian disruptions and CMDs were cross-sectional designs6, 11, 12, only very limited prospective analysis was conducted in large population13. Furthermore, advancements in sensor element technology and accelerometer now allow the objective measurement of human sleep-wake cycle, physical activity, and ambient environmental zeitgebers. Unlike biomarkers that objectively measured human circadian rhythm requiring repeated or invasive procedures14, 15, accelerometers enable a non-intrusive way to consistent tracking rest-activity cycles14-16. Notably, over 80% population in a large database (the Munich ChronotType Questionnaire database) reported using an alarm clock and accumulated sleep debts during working week17, indicating a large proportion of population suffer from misalignment between circadian preference and their actual sleep-wake cycle. However, little is known about whether and how the misalignment between circadian preference and actual sleep-wake cycle is associated with CMDs, particularly in large prospective cohort studies. Thus, based on a large-scale cohort study, the UK Biobank, we used the information on accelerometer-derived sleep-wake cycle, the self-reported circadian preference, and comprehensive diseases diagnosis resources 1) to prospectively analyze the association of circadian misalignment with CMDs such as T2D, CHD, and stroke, and 2) to reveal the potential mechanism underlying the relationship between circadian misalignment and CMDs, through mediation analysis by plasma biomarkers and anthropometric indices. ## Methods ### Study design and participants The UK Biobank, initiated in 2006-2010, is a large-scale prospective cohort study containing comprehensive health information from over half a million adult UK participants. The UK Biobank was granted ethical approval from the North West Multi-center Research Ethical Committee (reference # 11/NW/0382) and research is performed in accordance with the Declaration of Helsinki. This research has been conducted using the UK Biobank Resource under application number 96083. Among 502,505 participants at baseline in the UK Biobank, 74,123 participants with complete and valid information on both baseline sleep questionnaire and accelerometer derived information were included, after further excluding 7,052 participants with prevalent diabetes, stroke, or CHD at the time of monitoring activity, 389 participants with baseline random glucose level ≥ 11.1 mmol/L or HbA1c ≥ 48 mmol/mol, 5,418 participants reported jobs involves shift work, and 299 participants with measured sleep duration < 4 hours or > 12 hours, 60,965 participants were included in final analysis (**Figure 1**). ![Figure 1.](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2024/06/28/2024.06.28.24309628/F1.medium.gif) [Figure 1.](http://medrxiv.org/content/early/2024/06/28/2024.06.28.24309628/F1) Figure 1. Flow chart of the study ### Sleep assessment Sleep status was evaluated through both subjective touchscreen sleep questionnaires and objective actigraphy monitors. #### Sleep questionnaire At baseline recruitment, information on personal chronotype, 24h sleep duration, snoring, insomnia, and daytime sleepiness was collected by a self-administrated touchscreen sleep questionnaire. They were collected by the following questions: “Do you consider yourself to be? (definitely a ‘morning’ person, more a ‘morning’ person than an ‘evening’ person, more an ‘evening’ person than a ‘morning’ person, definitely an ‘evening’ person, do not know, prefer not to answer)”, “About how many hours sleep do you get in every 24 hours? (please include naps)”, “Does your partner or a close relative or friend complain about your snoring? (yes, no, do not know, prefer not to answer)”, “Do you have trouble falling asleep at night or do you wake up in the middle of the night? (never/rarely, sometimes, usually, prefer not to answer)”,” How likely are you to doze off or fall asleep during the daytime when you don’t mean to? (e.g., when working, reading or driving) (never/rarely, sometimes, often, do not know, prefer not to answer, all of the time)”. For all the above questions, answers with “do not know”, “prefer not to answer” were considered as missing. Healthy sleep quality score was calculated from questionnaire derived chronotype, insomnia, snore, daytime sleepiness, and sleep duration, see **Supplementary TableS1.** #### Actigraphy During June 2013 and January 2016, physical activity and sleep measurements were recorded via a wrist-worn accelerometer (Axivity AX3), which is a commercial version of the Open Movement AX3 open-source sensor ([https://github.com/digitalinteraction/openmovement](https://github.com/digitalinteraction/openmovement)) designed by Open Lab, Newcastle University. Exquisite processes on data collection and curation have been described previously18. Briefly, participants were instructed to wear the accelerometer upon receipt and proceed with their usual routines, affixing it to their dominant wrist. They were notified that the device would activate automatically upon delivery and deactivate after a week. At the end of this period, participants returned the device to the central office using a provided pre-paid envelope. R package GGIR was used to clearing, calibrating, and deriving information on timing of sleep, moderate to vigorous physical activity, and light exposure19, 20. ### Ascertainment of circadian misalignment The morning-evening preference, chronotype, was collected by one validated question from the Horne and Östberg Morningness-Eveningness Questionnaire (MEQ)21. This question indicated a good correlation with the overall score of the MEQ (r = 0.72)22, and was also employed in other large cohort studies to indicate individual circadian preference13, 23. Additionally, information on participants timing of onset and end of sleep was collected by wearing an actigraphy on dominant wrist through seven consecutive days and derived by R package GGIR. Nighttime sleep duration was calculated as the difference between time point of sleep end and sleep onset. Midsleep of weekdays, weekend, and weekly average can be calculated using the following formula, using the average value of sleep onset and sleep duration on weekdays, weekend, and the whole week, respectively. ![Formula][1] The misalignment between circadian preference (chronotype) and midsleep was calculated as the residuals of midsleep on chronotype or the difference between midsleep and chronotype (subtracting group numbers of chronotype from midsleep)24, detailed calculation and categorization can be found in **Supplementary Method**. ### Cardiometabolic outcomes Newly onset of T2D, CHD, and stroke (including ischemic stroke and hemorrhage stroke subtypes) were individually evaluated and collectively considered when calculating CMD incidence. Data on data showcase-first occurrence (Category 1712) were used to derive the first occurrence date of diseases. Two main classification systems of clinical coding were used in the linked health data: International Classification of Diseases (ICD) and Read. Information on primary cares (Read v2 and Read CTV3), hospital inpatient (ICD10 and ICD9), death register (ICD10), cancer register (ICD10 and ICD9), and self-reported medical condition were gathered and mapped to 3-character ICD-10 to ascertain the source and date of the first occurrence of diseases. Detailed method can be found here ([https://biobank.ndph.ox.ac.uk/showcase/ukb/docs/first\_occurrences_outcomes.pdf](https://biobank.ndph.ox.ac.uk/showcase/ukb/docs/first_occurrences_outcomes.pdf)). ICD code and field ID of outcomes were described in **Supplementary table S2**. Incident cases were defined as cases with a later date of first occurrence in comparison with the date of measurement of baseline information and accelerometer. ### Covariates Baseline information on demographic and socioeconomic status (age, sex, education, Townsend deprivation index, region, ethnic group, employment status) and family history of diseases were collected in assessment center when participants were initially recruited. Lifestyle factors including drinking and smoking status, tea and coffee intake, as well as a detailed 29-item food frequency questionnaire on dietary intakes were also evaluated in assessment center by touchscreen questionnaire. Diet intake information was converted to a Mediterranean diet score to indicate dietary quality25. Additionally, baseline body weight, height, waist circumference, hip circumference, and blood pressure were measured in assessment center under standard process. Body mass index was calculated as weight (kg) divided by height (m) squares, and waist-hip-ratio was also calculated. Blood samples were collected at recruitment, separated by components and stored at UK Biobank (−80°C and LN2). Sample collection and processing has been previously described in detail ([https://biobank.ndph.ox.ac.uk/showcase/ukb/docs/biomarker_issues.pdf](https://biobank.ndph.ox.ac.uk/showcase/ukb/docs/biomarker_issues.pdf)). A range of key biochemistry markers (hepatic and renal function, inflammatory cytokines, lipids and glucose metabolism) were used in the current analysis. Levels of glucose and HbA1c at baseline were employed to identify participants with prevalent diabetes. Levels of Low-Density Lipoprotein was also used in analysis, as it was closely related to cardiometabolic outcomes26. ### Statistical analysis All statistical analyses were performed using SAS 9.4 (SAS Institute Inc) and R version 3.4.2. Two-sided *P*<0.05 were regarded as the level of significance. Descriptive data of participants at baseline were shown as number (%) for categorical variables and mean (standard deviation) for continuous variables. See **Supplementary Method** for the adjusted circadian misalignment according to baseline characteristics and selection of covariates in the following models. Cox proportional hazard models were used to investigate the associations between circadian misalignment (quintiles of residuals of midsleep on chronotype) and incident cardiometabolic events. All cox models have passed the proportional hazards assumption by Schoenfield Residuals Test. Model 1 was adjusted for age (tertiles), sex (men/women), and ethnic group (White British/Non-White British), and Model 2 was additionally adjusted for region (urban/town/village), Townsend deprivation index (tertiles), employment (in-paid employment/retired/without fixed employment), qualification (college or university degree/A levels or equivalent/O levels or GCSEs or equivalent/none of the above), and ambient noise level (tertiles). Model 3 (main model) was further adjusted for smoking status (never/ever/current), drinking status (never/ever/current), tea consumption (<2/2-4/≥5 cups), coffee consumption (0/0-2/≥3 cups), accelerometer-derived moderate to vigorous physic activity (tertiles), and healthy sleep quality score (<2/2-3/>3). All continuous covariates were converted to tertiles to clearer show their relationship with circadian misalignment (Figure S1) and CMDs. Since missing values for all covariates were less than 6% (5.9% for healthy sleep quality score, and 1.2% for ambient noise level, and less than 0.2% for other covariates), they were imputed as the mode12. Plots on restricted cubic splines (four knots on 0.05, 0.35, 0.65, and 0.95, respectively) and *P* for nonlinear (ANOVA test27) were used to show the nonlinear relationship between circadian misalignment and cardiometabolic events. Moreover, risk of cardiometabolic events were also evaluated by circadian misalignment using difference between midsleep and chronotype (advanced/intermediate/delayed group), as well as chronotype and midsleep individually. When investigating the association of midsleep and chronotype with CMDs, they were further mutually adjusted. Sensitivity analyses were conducted by 1) recategorizing the main exposure, circadian misalignment (residuals of midsleep on chronotype), into three groups [advanced (Q1), intermediate (Q2+Q3+Q4), delayed (Q5) group]; 2) recalculating the main exposure, circadian misalignment, as the absolute value of the residuals of midsleep on chronotype, and categorized it into tertiles. The linear trend of HRs over tertiles was assessed by *χ2* test using these tertile numbers as continuous variables; 3) based on main model (Model 3), additionally adjusted for family history of diabetes (when outcome is T2D), family history of cardiovascular diseases (when outcome is stroke and CHD), or family history of diabetes and cardiovascular diseases (when outcome is CMDs), and time span between baseline recruitment and accelerometer assessment, as they were suggested confounders in previous studies. Age was adjusted in the form of continuous variable to reduce potential residual confounding; 4) Based on main model, further adjusted for BMI, LDL, and SBP, as they may act both as underlying mediators and confounders; 5) excluding upper and lower 0.5% value of residuals of midsleep on circadian preference; 6) calculating propensity score to balance the distribution of baseline characteristics among differed circadian misalignment group. Propensity score weights for multiple treatments and average treatment effect on the treated was calculated by the R package Twang and using the method of gradient boosting machine29, 30; 7) using ICD-10 code E11 instead of pooling E11 and E14 together to define T2D; and 8) including participants who worked with shift work in analysis, as shift work was an important contributor of circadian misalignment. In secondary analysis, causal mediation analyses were conducted by R package CMAverse28. Average total effect hazard ratio, nature direct hazard ratio, nature indirect hazard ratio, and proportion mediated by potential mediators were shown. Mediators were log-transformed and standardized to show HRs (95% CI) per standard deviation. Proportional hazards assumptions were also confirmed. See **Supplementary Method** for the selection of potential mediators. Furthermore, subgroup analyses were conducted by age (<57 /≥57 years (median)), sex (women/men), ethnic (White British/non-White British), residence (Urban/town or village), socioeconomic status (