Sleep Regularity and Mortality: A Prospective Analysis in the UK Biobank

Background Irregular sleep-wake timing may cause circadian disruption leading to several chronic age-related diseases. We examined the relationship between sleep regularity and risk of all-cause, cardiovascular disease (CVD), and cancer mortality in 88,975 participants from the prospective UK Biobank cohort. Methods The sleep regularity index (SRI) was calculated as the probability of an individual being in the same state (asleep or awake) at any two time points 24 hours apart, averaged over 7-days of accelerometry (range 0–100, with 100 being perfectly regular). The SRI was related to the risk of mortality in time-to-event models. Findings The mean sample age was 62 years (SD, 8), 56% were women, and the median SRI was 60 (SD, 10). There were 3010 deaths during a mean follow-up of 7.1 years. Following adjustments for demographic and clinical variables, we identified a non-linear relationship between the SRI and all-cause mortality hazard (p [global test of spline term] < 0·001). Hazard Ratios, relative to the median SRI, were 1·53 (95% confidence interval [CI]: 1·41, 1·66) for participants with SRI at the 5th percentile (SRI = 41) and 0·90 (95% CI: 0·81, 1·00) for those with SRI at the 95th percentile (SRI = 75), respectively. Findings for CVD mortality and cancer mortality followed a similar pattern. Conclusions Irregular sleep-wake patterns are associated with higher mortality risk. Funding National Health and Medical Research Council of Australia (GTN2009264; GTN1158384), National Institute on Aging (AG062531), Alzheimer’s Association (2018-AARG-591358), and the Banting Fellowship Program (#454104).


Table of Contents
Appendix 1: Removal of low-quality accelerometer data Appendix 2: STROBE checklist Table S1: Correlation between sleep regularity index and standard deviation-based regularity metrics Figure S1. Directed acyclic graph for identification of adjustment variables Figure S2. Time-varying HRs for 5 th and 95 th percentiles of SRI (relative to median) for all-cause mortality Figure S3. Time-varying HRs for 5 th and 95 th percentiles of SRI (relative to median) for cancer-mortality CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 15, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 Appendix 1: Methods. Removal of low-quality accelerometer data Accelerometry data of low quality were removed using established UKB criteria; incongruity of self-reported wear time and accelerometer wear time data (5%); insufficient wear time (< 72 hours; 5%); and poorly calibrated data (<1%). Lastly, data were removed for participants in which GGIR was unable to determine a sleep window (5%) and for participants providing less than two valid SRI measurements (i.e., 2 24-hour wear periods; <1%). In total, 88,975 (84%) participants provided valid sleep regularity index data and were included in the study.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 15, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 Introduction Background/rationale 2 Explain the scientific background and rationale for the investigation being reported (pg 3) Objectives 3 State specific objectives, including any prespecified hypotheses (pg 3)

Study design 4
Present key elements of study design early in the paper (Pg 3) Setting 5 Describe the setting, locations, and relevant dates, including periods of recruitment, exposure, follow-up, and data collection (Pg 3) Participants 6 (a) Give the eligibility criteria, and the sources and methods of selection of participants. Describe methods of follow-up (Pg 3-4) (b) For matched studies, give matching criteria and number of exposed and unexposed NA Variables 7 Clearly define all outcomes, exposures, predictors, potential confounders, and effect modifiers. Give diagnostic criteria, if applicable (Pg 4) Data sources/ measurement 8* For each variable of interest, give sources of data and details of methods of assessment (measurement). Describe comparability of assessment methods if there is more than one group (Pg 3-4) Bias 9 Describe any efforts to address potential sources of bias (pg 5) Study size 10 Explain how the study size was arrived at (pg 3 and appendix) Quantitative variables 11 Explain how quantitative variables were handled in the analyses. If applicable, describe which groupings were chosen and why (pg 5) Statistical methods 12 (a) Describe all statistical methods, including those used to control for confounding (pg 4-5, Figure S1) (b) Describe any methods used to examine subgroups and interactions NA (c) Explain how missing data were addressed (pg 4-5) (d) If applicable, explain how loss to follow-up was addressed (e) Describe any sensitivity analyses (pg 5)

Results
Participants 13* (a) Report numbers of individuals at each stage of study-eg numbers potentially eligible, examined for eligibility, confirmed eligible, included in the study, completing follow-up, and analysed (Pg 3) (b) Give reasons for non-participation at each stage (c) Consider use of a flow diagram Not considered necessary but can be created upon request Descriptive data 14* (a) Give characteristics of study participants (eg demographic, clinical, social) and information on exposures and potential confounders . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 15, 2023. ; https://doi.org/10.1101/2023.04.14.23288550 doi: medRxiv preprint Other analyses 17 Report other analyses done-eg analyses of subgroups and interactions, and sensitivity analyses (pg 6)

Discussion
Key results 18 Summarise key results with reference to study objectives (pg 7) Limitations 19 Discuss limitations of the study, taking into account sources of potential bias or imprecision. Discuss both direction and magnitude of any potential bias (pg 8) Interpretation 20 Give a cautious overall interpretation of results considering objectives, limitations, multiplicity of analyses, results from similar studies, and other relevant evidence (pg 8) Generalisability 21 Discuss the generalisability (external validity) of the study results (pg 7-8)

Other information
Funding 22 Give the source of funding and the role of the funders for the present study and, if applicable, for the original study on which the present article is based (pg 8) *Give information separately for exposed and unexposed groups. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 15, 2023. ; https://doi.org/10.1101/2023.04.14.23288550 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 15, 2023. ; https://doi.org/10.1101/2023.04.14.23288550 doi: medRxiv preprint Figure S1. Directed acyclic graph for identification of adjustment variables. The green node indicates the exposure variable (SRI), and the blue node (Mortality) is the outcome variable. Pale grey nodes indicate unobserved variables; white nodes indicate a variable which has been conditioned on (by regression adjustment or restriction). Paths in red are biasing paths. Arrows indicate the direction of causal effect between two nodes. P is an unobserved variable representing unmeasured causes of sleep habits (e.g., genetics). U is an unobserved variable representing unmeasured causes of disease and cardiovascular dysfunction (e.g., genetics, biological ageing). Z is an unobserved variable representing unmeasured causes of health behaviours (e.g., personality factors, genetics). Green paths from SRI to Prevalent disease, BP medication, Systolic BP, BMI, and Physical activity and from these nodes to Mortality represent potential mediation of an SRI effect. Conversely, red paths indicate potential sources of confounding (e.g., a backdoor path from Mortality to Prevalent disease to SRI via U). Given the current evidence base, we are unable to determine whether and to what extent variables such as Prevalent disease act as mediators or confounders (via U) of the SRI-mortality association. AP = anti-psychotic; AD = antidepressant; BMI = body mass index; BP = blood pressure; CVD = cardiovascular disease; Deprivation = the Townsend deprivation index; SRI = sleep regularity index; WASO = wake after sleep onset.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 15, 2023. ; https://doi.org/10.1101/2023.04.14.23288550 doi: medRxiv preprint Figure S2. Time-varying HRs for 5 th and 95 th percentiles of SRI (relative to median) for all-cause mortality. A: Hazard ratios for 5 th percentile vs median SRI; B: Hazard ratios for 95 th percentile vs median SRI. Discrete time hazards model including time (aggregated into 3-month intervals and modelled with a restricted cubic spline with knots at the 5 th , 35 th , 65 th , and 95 th percentiles), SRI, and an SRI by time interaction. Adjusted for age, Townsend deprivation index, sex, antidepressant, antipsychotic, and sedative medication, ethnicity, household income, education, smoking status (former, current, never), smoking pack years, shift work, retirement status, and sick or disabled (self-reported employment category). All continuous confounders and the SRI were modelled with restricted cubic splines (knots at 10 th , 50 th , and 90 th percentiles) to allow for departures from linearity. There was strong evidence of an interaction between time and SRI (p [interaction] < 0.001).
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 15, 2023. ; https://doi.org/10.1101/2023.04.14.23288550 doi: medRxiv preprint Figure S3. Time-varying HRs for 5 th and 95 th percentiles of SRI (relative to median) for cancer-mortality. A: Hazard ratios for 5 th percentile vs median SRI; B: Hazard ratios for 95 th percentile vs median SRI. Discrete time hazards model including time (aggregated into 3-month intervals and modelled with a restricted cubic spline with knots at the 5 th , 35 th , 65 th , and 95 th percentiles), SRI, and an SRI by time interaction. Adjusted for age, Townsend deprivation index, sex, antidepressant, antipsychotic, and sedative medication, ethnicity, household income, education, smoking status (former, current, never), smoking pack years, shift work, retirement status, and sick or disabled (self-reported employment category). All continuous confounders and the SRI were modelled with restricted cubic splines (knots at 10 th , 50 th , and 90 th percentiles) to allow for departures from linearity. There was strong evidence of an interaction between time and SRI (p [interaction] < 0.001).
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 15, 2023. ; https://doi.org/10.1101/2023.04.14.23288550 doi: medRxiv preprint Figure S4. SRI and CVD-specific mortality by sex. Adjusted for age, Townsend deprivation index, antidepressant, antipsychotic, and sedative medication, ethnicity, household income, education, smoking status (former, current, never), smoking pack years, shift work, retirement status, and sick or disabled (self-reported employment category). Hazard ratios are relative to the median SRI (SRI = 60).
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 15, 2023. ; https://doi.org/10.1101/2023.04.14.23288550 doi: medRxiv preprint Figure S5. SRI and all-cause mortality in sensitivity analyses. P values from global (2 degree of freedom) test of spline term. Hazard ratios are relative to the median SRI (SRI = 60).
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 15, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 Figure S6. SRI and CVD-mortality in sensitivity analyses. P values from global (2 degree of freedom) test of spline term. Hazard ratios are relative to the median SRI (SRI = 60).
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 15, 2023. ; https://doi.org/10.1101/2023.04.14.23288550 doi: medRxiv preprint Figure S7. SRI and cancer-mortality in sensitivity analyses. P values from global (2 degree of freedom) test of spline term. Hazard ratios are relative to the median SRI (SRI = 60).
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 15, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023