What this is
- This research investigates the connection between sleep behaviors and the risk of developing rheumatoid arthritis (RA) using data from the UK Biobank.
- It evaluates how various sleep factors, including sleep duration and insomnia, influence RA risk, particularly in individuals with different genetic predispositions.
- The study also explores the combined effects of sleep and genetic factors on RA risk, highlighting significant findings from a large cohort.
Essence
- Poor sleep patterns, particularly short or long sleep duration and insomnia, are linked to a higher risk of rheumatoid arthritis (RA). Participants with a high () had a 70.2% increased risk of RA compared to those with a low .
Key takeaways
- Short sleep duration (less than 7 hours) and long sleep duration (more than 8 hours) are associated with a higher risk of RA. Hazard ratios indicate a 1.389Γ and 1.482Γ increased risk for these groups, respectively.
- Insomnia significantly increases RA risk, with those experiencing frequent insomnia having a 1.628Γ higher risk compared to those who never or rarely experience it.
- Participants with high scores had a 70.2% increased risk of developing RA compared to those with low scores, indicating a strong association between poor sleep and RA risk.
Caveats
- The study's findings are based on a European Caucasian population, which may limit generalizability to other racial groups. Additionally, sleep behaviors were assessed only at baseline, which does not account for changes over time.
- Self-reported sleep data could introduce measurement errors, and despite adjustments for known confounders, unmeasured factors may still influence the results.
Definitions
- Poor Sleep Score (PSS): A composite measure ranging from 0 to 5 that assesses sleep quality based on multiple sleep behaviors, with higher scores indicating poorer sleep.
- Genetic Risk Score (GRS): A score calculated based on specific genetic variants associated with RA, used to categorize individuals into low, intermediate, or high genetic risk groups.
AI simplified
Introduction
Rheumatoid arthritis (RA) is a chronic systemic inflammatory disease characterized by synovitis. If left untreated, it can lead to joint deformity and loss of function. The pathogenesis of RA is still under investigation, but it may involve a combination of autoimmune factors, genetics, smoking, and other environmental influences [1]. Individuals with RA often experience pain, obesity, and poor sleep, which significantly reduce their quality of life compared to those without the condition [2,3].
Sleep is essential for maintaining human health [4,5]. While the relationship between RA and sleep has been extensively studied, it remains unclear whether poor sleep is a cause or consequence of RA. In most cases, sleep disturbances in RA patients are linked to pain. A Swedish case-control study found that sleep problems worsen as the duration of the disease is extended [6]. A cross-sectional, multicenter study involving 305 Belgian patients reported that poor sleep quality is associated with inadequate RA control, potentially due to pain [5]. Moreover, medications prescribed for RA treatment may further impact sleep quality [5,7,8]. Recent research has highlighted a strong connection between sleep and the immune system, which plays a crucial role in complex diseases [9]. During sleep, the immune system releases proteins called cytokines, some of which promote sleep. In the presence of infection or inflammation, the demand for certain cytokines increases [10]. Conversely, sleep deprivation may reduce the production of these protective cytokines [9], leaving the body with fewer antibodies and cells to combat infection [11].
The relationship between poor sleep and various complex diseases has been a focus of recent investigation. Studies have shown that women with extreme sleep durations are more likely to develop metabolic syndrome and exhibit higher metabolic syndrome severity scores. In contrast, women who consistently sleep for seven hours per night have the lowest severity scores for metabolic syndrome [12]. A meta-analysis revealed a U-shaped relationship between sleep duration and the risk of osteoporosis, with the lowest risk observed among individuals who slept for eight hours per night [13]. Additionally, a UK cohort study demonstrated that subjects who slept fewer than four hours or more than ten hours per night experienced significantly faster cognitive decline compared to those who slept seven hours per night [14]. Either too much or too little sleep increased all-cause mortality and cardiovascular events, with 7 hours of sleep per night having the lowest risk [15]. Several studies have also suggested that poor sleep might contribute to the development of RA [16β18]. However, most of these studies are limited by cross-sectional designs and small sample sizes, and have only established a correlation between sleep and RA. There is a lack of research on whether sleep characteristics modulate the relationship between genetic factors and RA. Importantly, the interaction between sleep characteristics and genetic factors in determining RA risk remains largely unknown.
A composite measure of sleep health, known as the sleep score, was recently proposed. A prospective study examined the combined association of five sleep behaviorsβsleep duration, chronotype, insomnia, snoring, and excessive daytime sleepinessβin the form of a healthy sleep score, with the risk of cardiovascular disease (CVD) [19].
Building on this concept, we prospectively assessed the effects of seven sleep behaviors on the risk of developing RA. To better understand the combined effect of poor sleep patterns and high genetic risk on RA, we followed [19] and constructed a poor sleep score (PSS) based on key sleep behaviors. Additionally, we examined the mediation effect of inflammation as a potential pathway from PSS to RA incidence. Furthermore, we explored the combined influence of sleep characteristics and genetic susceptibility on RA risk, and investigated potential gene-environment interactions. Finally, we conducted a case-control study using baseline data to investigate the impact of RA on sleep. The findings of our research are illustrated in Fig 1.
Overflow of our research.
Materials and methods
Study design
We first examined the associations between multiple sleep behaviors and RA risk in a prospective study, calculating the PSS for each participant to evaluate its effect on RA risk. The participants were then stratified based on their calculated sleep scores and Genetic Risk Scores (GRS) to investigate the interaction and combined effects of sleep and genetic factors on RA risk. Additionally, a case-control study was conducted to assess the impact of RA onset on sleep quality in later life. Finally we used linkage disequilibrium scores regression (LDSC) [20] to assess the genetic association between sleep behaviors and RA.
UK Biobank data sets
The UK Biobank is a large biomedical database initiated in 2006, containing comprehensive genetic and health data from 500,000 UK participants aged 40 to 70 (http://www.ukbiobank.ac.uk/β). The study has collected, and continues to collect, detailed genotypic and phenotypic information, including questionnaires, physical measurements, multimodal imaging, and data for a wide range of health-related longitudinal follow-up investigations [21]. Participants from 22 assessment centers with each eligible participant completing a written informed consent form were included at baseline. This database offers extensive health-related information, including biological measurements, lifestyle indicators, and biomarkers in blood and urine. It provides researchers with a unique opportunity to explore the combined effects of genes and the environment on complex diseases, offering a novel approach to understanding disease development. The present analyses were conducted under UK Biobank application number 76875. Detailed information is as follows:
Ethics approval statement.
Ethical approval was granted by the Northwest Centre for Research Ethics Committee, and informed consent was obtained from all participants.
Definition of RA cases.
Rheumatoid arthritis (RA) was defined according to the International Classification of Diseases, Tenth Revision (ICD-10-CM), as recorded in the UK Biobank (coded as βM05β and βM06β), with seropositive RA defined by βM05.β The outcome event was determined by the onset of RA from baseline to the date of diagnosis, death, loss to follow-up, or January 1, 2022, whichever occurred first. Self-reported non-cancer illness (coded as β1524β) was used to identify RA cases at baseline, which were included in case-control studies and excluded from prospective studies.
Definition of sleeping condition.
The data on sleeping habits of the participants was obtained through detailed self-reported questionnaires collected by the UK Biobank. A total of seven sleeping phenotypes were included in our study: sleep duration (Field ID: 1160), difficulty getting up (Field ID: 1170), chronotype (Field ID: 1180), nap during the day (Field ID: 1190), sleeplessness/insomnia (Field ID: 1200), snoring (Field ID: 1210), daytime dozing/sleeping (Field ID: 1220). The definitions of these sleep traits are provided in S1 Table. Poor sleep factors were defined as follows: late chronotype (βeveningβ or βevening than morningβ); sleep less than 7h or more than 8h per day; reported βsometimesβ or βusuallyβ insomnia symptoms; self-reported snoring; and frequent daytime sleepiness (βusuallyβ). The PSS was generated, incorporating five sleep factors (chronotype, duration, insomnia, snoring, and excessive daytime sleepiness) to better assess an individualβs sleep status and to accurately classify the population. The PSS ranges from 0 to 5, with higher scores indicating a poorer sleep pattern. Sleep patterns were categorized as βhealthy sleep patternβ (PSS β€ 1), βintermediate sleep patternβ (2 β€ PSS β€ 3), and βpoor sleep patternβ (PSS β₯ 4) based on the constructed PSS. Additionally, a weighted PSS (PSSCV) was created using a 10-fold cross-validation procedure as a sensitivity analysis. The weighted PSS was derived by combining the five sleep factors.
Summary-level data and construction of genetic risk score (GRS)
Genetic data on sleep behaviors (sleep duration [22], difficulty getting up [23], chronotype [24], nap during day [25], insomnia [26], snoring [27], daytime dozing [28]) were utilized in the largest relevant genome-wide study of its kind to date.
In the genetic analysis, we randomly divided the total sample into a training set (20%) and a test set (80%). The training set was used to identify the best-performing GRS, and all subsequent analyses were conducted using the test set. The GRS was calculated based on genotypes of selected single nucleotide polymorphisms (SNP) from UK Biobank under different thresholds (P from 5 Γ 10β8 to 5 Γ 10β4), where LD r2 was set to 0.001 within 1Mb physical distance. The r2 was derived from the 1000 Genomes Projects for individuals of European ancestry. We applied strict quality control measures to both genotypes and individuals. For genotypes, variants were retained if the minor allele frequency (MAF) was greater than 0.02, the imputation information score was greater than 0.5, the missing genotype rate was less than 5%, and the Hardy-Weinberg equilibrium P-value was greater than 1 Γ 10 β» 7 [21]. The GRS for each individual is calculated by summing the product of genotypes (0, 1, 2) of risk loci and their corresponding weights, which were from the largest GWAS analysis for RA of European populations [29]. Additionally, we normalized the GRS to a mean of 0 and a standard deviation of 1. We restricted the analysis to approximately 330,000 individuals with successful genotyping, of white British ancestry, and without identified relatives. The best-performing GRS was defined as the score with the strongest association with RA incidence. Finally, genetic risk groups were categorized based on the selected GRS: low genetic risk (bottom quartile), intermediate genetic risk (2nd and 3rd quartiles), and high genetic risk (top quartile). Detailed information on the SNPs used for the GRS is provided in our previous work [30,31].
Statistical analysis
The basic characteristics of the 379,314 participants were summarized using the mean (standard deviation) for continuous variables and frequencies (percentages) for categorical variables. A Cox regression model was employed to assess the effect of various sleep phenotypes on the risk of RA over time, calculating hazard ratios (HRs) and their corresponding 95% confidence intervals (CIs). Three models were constructed to examine the relationship between sleep behavior and RA risk in this observational study: Model 1, adjusted for genotyping batch (UKB vs. BiLEVE), assessment center, sex, Townsend Deprivation Index (TDI), age, and kinship; Model 2, adjusted for batch, assessment center, sex, TDI, age, kinship, body mass index (BMI), smoking status, and alcohol consumption; Model 3, adjusted for batch, assessment center, sex, TDI, age, kinship, BMI, smoking status, alcohol consumption, diet, physical activity (PA), diastolic blood pressure (DBP), systolic blood pressure (SBP), high-density lipoprotein (HDL) cholesterol, low-density lipoprotein (LDL) cholesterol, triglycerides, and cancer. The TDI is a regionally-based proxy for socioeconomic status. A restricted cubic spline curve and a linearity test were used to assess the non-linear relationship between continuous variables, such as sleep duration and the PSS, and the risk of RA. Incidence rates in different subgroups were evaluated in a prospective study, and Harrellβs c-index was used to evaluate model performance. To assess the potential mediated role of chronic inflammation between sleep and RA, we used an inflammation-based score, the modified Glasgow Prognostic Score (mGPS). Participants with both elevated serum C-reactive protein (CRP) (>1.0 mg/dL) and hypoalbuminemia (<3.5 g/dL) were assigned an mGPS of 2; those with one abnormal value were assigned an mGPS of 1, and those with neither abnormality were assigned an mGPS of 0 [32]. Covariates used in GRS analysis included age, sex, TDI, genotyping batch, and top 10 genetic principal components (PCs). Additionally, we plotted the cumulative incidences of RA based on genetic and sleep risk in this cohort.
In case-control research, propensity score matching (PSM) was used to select participants for the control group, adjusting for genotyping batch, assessment center, sex, TDI, age, race, BMI, smoking status, alcohol consumption, diet, and physical activity. Linear regression was applied to examine the effect of RA onset on sleep duration, while ordinary logistic regression assessed the impact of RA onset on morning wakefulness, chronotype (morning/evening person), daytime napping, insomnia, daytime drowsiness, and PSS. Multivariable logistic regression was then used to evaluate the effect of RA onset on snoring.
Additionally, we plotted the cumulative incidences of RA according to genetic and sleep risk. Statistical significance was defined by a two-tailed P-value < 0.05. To account for multiple comparisons, Bonferroni correction was applied, with a significance threshold of P < 0.006 (0.05/8 sleep behaviors, including PSS). All statistical analyses were conducted using R (version 3.6.1) and PLINK software (version 1.90 b3.38) [33].
Results
Baseline characteristics
This study incorporated 379,314 subjects (comprising 171,207 men and 208,107 women), among whom 4,610 were diagnosed with RA. Table 1 shows the baseline characteristics of the study participants in the different PSS groups. Of all participants, 7.09%, 30.40%, 38.45%, 19.75% and 4.30% had a PSS of 0, 1, 2, 3 and 4β5, respectively (S2 Table). As PSS increased, the proportion of regular physical activity and a healthy diet decreased. In contrast, BMI was positively associated with PSS, while the risks of chronic kidney disease, type 2 diabetes, cancer, and RA increased with an increase in PSS.
| Traits | Level | Poor sleep score | ||
|---|---|---|---|---|
| Low score | Intermediate score | High score | ||
| Number of participants | 142,227 | 220,758 | 16,329 | |
| Age, years (mean (SD)) | 56.23 (8.24) | 56.85 (7.90) | 56.59 (7.78) | |
| Male (%) | 58235 (42.2) | 98157 (46.4) | 8243 (53.4) | |
| BMI (mean (SD)) | 26.49 (4.32) | 27.69 (4.78) | 29.59 (5.50) | |
| Townsend index (mean (SD)) | -1.76 (2.80) | -1.52 (2.96) | -1.02 (3.19) | |
| Tobacco smoking status (%) | Never | 81922 (59.4) | 108827 (51.4) | 6582 (42.6) |
| Previous | 45827 (33.2) | 78903 (37.3) | 6298 (40.8) | |
| Current | 10149 (7.4) | 23820 (11.3) | 2564 (16.6) | |
| Alcohol drinking status (%) | Never | 4595 (3.3) | 5990 (2.8) | 411 (2.7) |
| Previous | 4246 (3.1) | 6785 (3.2) | 665 (4.3) | |
| Current | 129015 (93.6) | 198682 (93.9) | 14359 (93.0) | |
| Regular physical activity (%) | 102946 (74.7) | 146836 (69.4) | 9497 (61.5) | |
| Healthy diet (%) | 51896 (37.6) | 69215 (32.7) | 4209 (27.3) | |
| Chronic kidney disease (%) | 3907 (2.8) | 7892 (3.7) | 834 (5.4) | |
| Type 2 diabetes (%) | 4780 (3.5) | 11477 (5.4) | 1564 (10.1) | |
| Cancer (%) | 21620 (15.7) | 38251 (18.1) | 3199 (20.7) | |
| RA (%) | 1387 (1.0) | 2903 (1.4) | 320 (2.1) | |
| Seropositive RA (%) | 170 (0.1) | 285 (0.1) | 34 (0.2) | |
| Sleep duration (mean (SD)) | 7.43 (0.62) | 7.04 (1.19) | 6.78 (1.75) | |
| Sleep duration (%) | 7β~β8h | 131092 (95.1) | 120523 (57.0) | 683 (4.4) |
| 7h- | 3964 (2.9) | 70356 (33.3) | 10746 (69.6) | |
| 8h+ | 2842 (2.1) | 20671 (9.8) | 4015 (26.0) | |
| The difficulty of getting up (%) | Very easy | 57851 (42.0) | 59560 (28.2) | 2172 (14.1) |
| Fairly easy | 68858 (49.9) | 106307 (50.3) | 6873 (44.5) | |
| Not very easy | 9709 (7.0) | 36045 (17.0) | 4417 (28.6) | |
| Not at all easy | 1480 (1.1) | 9638 (4.6) | 1982 (12.8) | |
| Chronotype (%) | Definitely morning | 50234 (36.4) | 45856 (21.7) | 558 (3.6) |
| Morning more | 72403 (52.5) | 59555 (28.2) | 501 (3.2) | |
| Evening more | 11784 (8.5) | 81704 (38.6) | 10391 (67.3) | |
| Definitely evening | 3477 (2.5) | 24435 (11.6) | 3994 (25.9) | |
| Nap during day (%) | Never/rarely | 84989 (61.6) | 117257 (55.4) | 6639 (43.0) |
| Sometimes | 47601 (34.5) | 82989 (39.2) | 6841 (44.3) | |
| Usually | 5308 (3.8) | 11304 (5.3) | 1964 (12.7) | |
| Sleeplessness (%) | Never/rarely | 64795 (47.0) | 24071 (11.4) | 159 (1.0) |
| Sometimes | 52841 (38.3) | 114064 (53.9) | 7257 (47.0) | |
| Usually | 20262 (14.7) | 73415 (34.7) | 8028 (52.0) | |
| Snoring (%) | No | 121672 (88.2) | 107238 (50.7) | 730 (4.7) |
| Yes | 16226 (11.8) | 104312 (49.3) | 14714 (95.3) | |
| Daytime dozing (%) | Never/rarely | 114064 (82.7) | 159619 (75.5) | 8500 (55.0) |
| Sometimes | 23541 (17.1) | 46437 (22.0) | 3569 (23.1) | |
| Usually | 293 (0.2) | 5494 (2.6) | 3375 (21.9) | |
Association between sleep behaviors and RA risk
Poor sleep conditions and RA risk.
In the prospective study, U-shaped relationships between sleep duration and RA risk were observed (Pnon-linearity < 0.001) by using RCS method in the primary analysis (Model 1), with 7h of sleep duration having the lowest risk of RA (Fig 2A). Similar U-shaped associations were observed in sex-stratified groups (Pnon-linearity < 0.001) (Fig 2B and 2C). Participants with short sleep duration (less than 7 hours) and long sleep duration (more than 8 hours) had a higher risk of RA than those who slept 7 to 8 hours per night after multiple test adjustments (HRsleep duration < 7 = 1.389, 95%CI = 1.303 ~ 1.480; HRsleep duration > 8 = 1.482, 95%CI = 1.352 ~ 1.624) (Table 2). After correcting for more covariates (e.g., BMI, physical activity, and diabetes), the association remained statistically significant in Models 2 and 3. An association between insomnia and RA risk was observed in Model 1. Compared to never/rarely insomnia, the HR estimate for sometimes insomnia was 1.168 (95%CI = 1.077 ~ 1.267) and for frequent insomnia was 1.628 (95%CI = 1.498 ~ 1.769). After correcting for more covariates, the association is barely weakened. In the fully adjusted model (model 3), usually insomnia was associated with incident RA with a 54.4% higher risk (HR: 1.544, 95% CI: 1.404 ~ 1.698). When compared to people who never or seldom encountered daytime dozing, those who usually experienced had a higher risk of developing RA, with the HRs (95% CI) estimated to be 1.683 (1.466 ~ 1.932) in model 1, which is significant after multiple test adjustments (P = 1.58 Γ 10β13). Associations were still significant after controlling for more covariates in Models 2 and 3. Besides, we found no evidence of an increased risk of RA due to chronotype and snoring.
Restricted cubic spline models for the relationship between sleep duration and RA risk. All participates;In men;In women. Note: Associations were adjusted for age, sex, genotyped batch, assessment center, kinship, TDI. Abbreviation: RA, rheumatoid arthritis; TDI, Townsend deprivation index. (A) (B) (C)
| Traits | Model 1 | Model 2 | Model 3 | ||||
|---|---|---|---|---|---|---|---|
| HR (95%CI) | P | HR (95%CI) | P | HR (95%CI) | P | ||
| Sleep duration, hours | 7β~β8 | Ref | Ref | Ref | |||
| <7 | 1.389 (1.303, 1.480) | 3.16βΓβ10-24 | 1.320 (1.238, 1.407) | 2.01βΓβ10-17 | 1.332 (1.240, 1.432) | 6.03βΓβ10-15 | |
| >8 | 1.482 (1.352, 1.624) | 3.28βΓβ10-17 | 1.382 (1.260, 1.515) | 6.56βΓβ10-12 | 1.285 (1.154, 1.429) | 4.35βΓβ10-6 | |
| Insomnia | Never/rarely | Ref | Ref | Ref | |||
| Sometimes | 1.168 (1.077, 1.267) | 1.81βΓβ10-4 | 1.169 (1.078, 1.269) | 1.77βΓβ10-4 | 1.168 (1.065, 1.282) | 9.74βΓβ10-4 | |
| Usually | 1.628 (1.498, 1.769) | 1.65βΓβ10-30 | 1.561 (1.435, 1.697) | 2.65βΓβ10-25 | 1.544 (1.404, 1.698) | 3.17βΓβ10-19 | |
| Chronotype | Definitely morning | Ref | Ref | Ref | |||
| Morning more | 0.921 (0.858, 0.989) | 0.024 | 0.955 (0.889, 1.026) | 0.211 | 0.957 (0.882, 1.039) | 0.294 | |
| Evening more | 1.017 (0.944, 1.095) | 0.661 | 1.005 (0.933, 1.084) | 0.888 | 1.010 (0.928, 1.100) | 0.811 | |
| Definitely evening | 1.087 (0.977, 1.209) | 0.127 | 1.016 (0.912, 1.132) | 0.769 | 1.031 (0.913, 1.163) | 0.625 | |
| Snoring | No | Ref | Ref | Ref | |||
| Yes | 1.155 (1.089, 1.225) | 1.45βΓβ10-6 | 1.038 (0.977, 1.102) | 0.229 | 1.028 (0.960, 1.100) | 0.429 | |
| Daytime dozing | Never/rarely | Ref | Ref | Ref | |||
| Sometimes | 1.226 (1.149, 1.309) | 9.39βΓβ10-10 | 1.185 (1.110, 1.266) | 4.03βΓβ10-7 | 1.168 (1.084, 1.259) | 4.70βΓβ10-5 | |
| Usually | 1.683 (1.466, 1.932) | 1.58βΓβ10-13 | 1.510 (1.312, 1.737) | 8.75βΓβ10-9 | 1.474 (1.257, 1.727) | 1.70βΓβ10-6 | |
| Nap during day | Never/rarely | Ref | Ref | Ref | |||
| Sometimes | 1.200 (1.131, 1.273) | 1.41βΓβ10-9 | 1.129 (1.064, 1.199) | 6.71βΓβ10-5 | 1.078 (1.007, 1.154) | 0.03 | |
| Usually | 1.487 (1.328, 1.665) | 5.79βΓβ10-12 | 1.318 (1.175, 1.478) | 2.35βΓβ10-6 | 1.358 (1.197, 1.540) | 1.88βΓβ10-6 | |
| The difficulty of getting up | Very easy | Ref | Ref | Ref | |||
| Fairly easy | 1.092 (1.022, 1.167) | 0.009 | 1.095 (1.025, 1.170) | 0.007 | 1.067 (0.990, 1.151) | 0.088 | |
| Not very easy | 1.560 (1.430, 1.702) | 1.02βΓβ10-23 | 1.509 (1.383, 1.647) | 2.67βΓβ10-20 | 1.484 (1.345, 1.637) | 3.77βΓβ10-15 | |
| Not at all easy | 2.372 (2.106, 2.670) | 3.27βΓβ10-46 | 2.131 (1.890, 2.403) | 5.26βΓβ10-35 | 1.870 (1.625, 2.152) | 2.62βΓβ10-18 | |
Daytime napping and difficulty of getting up and RA risk.
Similarly, naps during day and difficulty in getting up might also contribute to the development of RA. In Model 3, compared to those who never/rarely nap during day, participants who sometimes and usually nap had a 7.8% (HR: 1.078, 95%CI: 1.007β~β1.154) and 35.8% (HR: 1.358, 95%CI: 1.197β~β1.540) higher risk of developing RA. Interestingly, we also found that difficulty in getting up was also positively associated with the risk of RA, where subjects who got up not at all easy had around 2 times higher risk of RA than those who got up very easily in all models.
Sleep behaviors and seropositive RA risk.
We performed a sensitivity analysis to evaluate associations between sleeping conditions and seropositive RA risk. As shown in, two sleeping problems (i.e., long sleep duration and insomnia) were also associated with an increased risk of seropositive RA after multiple test adjustments. A long duration of sleep (sleeping β₯ 8h) was associated with an increased risk of seropositive RA. Usually insomnia have a higher risk of RA compared with never/rarely insomnia. Besides, difficulty getting up and daytime napping is also associated with an increased risk of seropositive RA. S3 Table
Association between poor sleep score (PSS) and RA risk
When a PSS was used to evaluate the combined effects of five sleep behaviors (i.e., sleep duration, insomnia, chronotype, snoring, and daytime dozing) on the risk of RA, we observed linear relationships between PSS and RA risk overall by using RCS method (Fig 3A), as well as sex-stratified groups (S1 Fig). As shown in S4 Table, participants with high PSS have a 70.2% (HR = 1.722, 95% CI: 1.487 ~ 1.949) increased risk of RA compared to those with low scores. Similar results were observed in the different PSS scores (S5 Table) and PSS calculated by cross-validation (PSSCV) (S2 Fig). The sensitivity analysis yielded consistent results that people with high PSSCV scores were 60.6% (HR = 1.606, 95%CI: 1.468 ~ 1.757) more likely to develop RA than those with low PSSCV scores. The mediation analyses showed that 2.47% of the effect of PSS on the risk of RA was mediated by mGPS (P = 0.002) (S3 Fig). PSS can increase mGPS levels and further increase RA risk (Ξ²PSS β mGRS = 0.020, Ξ²mGRS β RA risk = 0.197). We also observed an independent effect of PSS on RA risk after correction for mGPS (Ξ²Direct effect = 0.154). Linear relationships were also observed in associations between PSS and seropositive RA risk in overall and women (Pnon-linearity > 0.05) but not in men (Pnon-linearity = 0.038) (S3 Fig). As shown in S3 Table, the association between intermediate PSS and seropositive RA no longer existed. Similar results were observed in the different PSS scores (S4 Table) and PSSCV (S3 Table). The cumulative incidence of RA in different PSS groups was shown in Fig 3B.
Relationship between poor sleep score (PSS) and RA risk. Restricted cubic spline models for the relationship between PSS and RA risk;Standardized rates of RA events in low (bottom quartile), intermediate (quartiles 2 to 3), and high (top quartile) PSS groups in the UKB cohort;Standardized rates of RA events in participants with different genetic risk score groups and different PSS groups in the UKB cohort. Note: Associations inwere adjusted for age, sex, genotyped batch, assessment center, kinship, and TDI. Associations inandwere adjusted for genotype batch, assessment centre, sex, TDI, age, kinship, BMI, smoking status, alcohol, diet, PA, DBP, SBP, HDL cholesterol, LDL direct, Cholesterol, Triglycerides, and Cancer. Abbreviations: HR, hazard ratio; TDI, Townsend deprivation index; PA, physical activity; DBP, diastolic blood pressure; SBP, systolic blood pressure; HDL, high-density lipoprotein; LDL, low-density lipoprotein; PSS, poor sleep score. (A) (B) (C) A B C
The combined effect of sleeping traits and genetic factors on RA risk
After adjustment, we chose the GRS with P < 5 Γ 10β7 to represent the genetic component, which had the best performance in terms of the significance of associations (S6 Table). As shown in S7 Table, among all participants, with worsening insomnia, snoring, and daytime dozing, the risk of RA was significantly increased in the moderate GRS group and high GRS group, and participants with high GRS and the worst sleep quality had the highest risk of RA (HRinsomnia = 3.652, 95%CI = 2.752 ~ 4.846, HRsnoring = 2.306, 95%CI = 1.918 ~ 2.773, HRdaytime dozing = 3.805, 95%CI = 2.672 ~ 5.416). For sleep duration and chronotype, with a certain GRS, we also found a U-shaped relationship between sleep duration and RA risk, with the lowest risk observed for 7β8 hours of sleep per night (in the high GRS group, HRsleep duration7 ~ 8 = 2.109, 95%CI = 1.789 ~ 2.485, HRsleep duration < 7 = 3.158, 95%CI = 2.616 ~ 3.811, HRsleep duration > 8 = 2.888, 95%CI = 2.218 ~ 3.761), an evening person were more likely to suffer from RA than morning person (in the high GRS group, HRdefinitely morning = 2.045, 95%CI = 1.583 ~ 2.642, HRmorning more = 2.051, 95%CI = 1.606 ~ 2.619, HRevening more = 2.268, 95%CI = 1.764 ~ 2.916, HRdefinitely evening = 2.410, 95%CI = 1.741 ~ 3.336). Furthermore, given the same sleep conditions, we found that the high GRS group had a higher risk of RA than the medium GRS group, whereas no such results were observed in the low GRS group (except for sleep duration). In the sex-stratified analysis, the combined effect of sleep factors and GRS was greater in women than in men.
We next assessed the combined association of PSS and GRS on the risk of developing RA, detailed results were presented in Fig 4. According to our findings, the combined effect of GRS and PSS enhanced the chance of RA incident. We also discovered that subjects with high PSS and high GRS had the highest risk of developing RA (HR = 4.062, 95%CI: 2.792 ~ 5.910). Results stratified by gender showed that women were more likely than men to be affected by the combined effects of PSS and GRS. In the high PSS and high GRS group, the HRs (95% CI) were estimated to be 3.880 (2.230 ~ 6.750) for men, and 4.072 (2.440 ~ 6.797) for women. Similar results were also observed with the PSS calculated by using cross-validation method (S4 Fig). Fig 3C shows the cumulative incidence of RA events in participants with different GRS groups and different PSS groups. C-statistic was calculated to be 0.657 in model without GRS and 0.672 in model with GRS (2.2% improvement).
RA risks in the subgroups stratified by genetic risk and poor sleep score (PSS) (versus participants with the lowest PSS in the lowest genetic risk group) in the UKB cohort. Note: Associations were adjusted for age, sex, TDI, genotyping batch, and top 10 genetic PCs. Abbreviations: PSS, poor sleep score; CI, confidence interval; HR, hazard ratio; PC, principal components; GRS, genetic risk score.
Effect of RA onset on sleep condition
As shown in S8 Table, in the case-control study, there were statistically significant associations between RA onset and poor sleep conditions (i.e., PSS) (Ξ² = 0.191, 95%CI = 0.095 ~ 0.288). Besides, ordinal regression analysis indicated that RA onset could increase the risk of difficulty getting up (OR = 1.293, 95%CI = 1.156 ~ 1.446,). However, no significant associations were observed between RA onset and other sleep behaviors, including insomnia, sleep duration, and daytime drowsiness, in later life.
Genetic correlations between RA and sleep phenotypes
In the analysis of genetic correlations using summary data from GWAS, we observed positive genetic associations between RA and insomnia (rg = 0.13; P = 0.0002) and chronotype (rg = 0.1; P = 0.00008), as well as negative genetic associations with sleep duration. Similar results were observed in seropositive RA (S5 Fig provides detailed information).
Discussion
This study, utilizing large-scale data from the UK Biobank, this study systematically explored the relationship between seven sleep behaviors and RA risk. The prospective study indicated that abnormal sleep duration, insomnia, and daytime dozing could increase the risk of RA. A U-shaped relationship between sleep duration and RA risk, with the lowest risk of RA when the daily sleep duration was 7β8 hours. Additionally, a PSS was developed to evaluate the combined effects of five poor sleep behaviors, revealing that higher PSS values were linked to an increased RA risk. In particular, individuals in the highest PSS group were found to have a 70.2% greater likelihood of developing RA. We also conducted a mediation analysis, which revealed that the mGPS acted as a mediator between sleep behaviors and RA risk. Furthermore, we observed the combined effects of genetic factors and PSS on RA risk, with sex-stratified analysis showing a stronger effect in women than in men. The case-control study further confirmed that RA risk is associated with long sleep duration, insomnia, and difficulty waking up in the morning. Lastly, significant genetic correlations were identified between sleep phenotypes and various types of RA.
Our findings provide robust evidence supporting the association between poor sleep and an elevated risk of RA, particularly with abnormal sleep duration, insomnia, and daytime drowsiness. A study conducted in China similarly found that poor sleep quality and short sleep duration increase RA risk [34], consistent with our results. However, we identified a U-shaped relationship between sleep duration and RA risk, with the lowest risk found in individuals who slept 7β8 hours per day. This pattern is consistent with studies on other diseases and sleep duration. For instance, research by Abbas Smiley et al. showed that women with extreme sleep durations had an increased likelihood of developing metabolic syndrome, with the lowest severity scores observed among those sleeping 7 hours per night [12]. Similarly, a meta-analysis reported a U-shaped relationship between sleep duration and osteoporosis risk, with the lowest risk for 8 hours of sleep per night [13]. A cohort study of 65,754 patients in Taiwan found a link between sleep disorders and a higher risk of RA [16], and a study by a conclusion supported by Taiwanβs National Health Insurance Scheme [17]. A UK cohort study revealed that faster cognitive decline in individuals who slept fewer than 4 hours or more than 10 hours per night, compared to those who slept 7 hours [14]. Jiawei Yin et al. also demonstrated that both insufficient and excessive sleep increase all-cause mortality and cardiovascular events, with 7 hours of sleep per night associated with the lowest risk [15]. Notably, we also observed significant associations between poor sleep patterns and RA risk in seropositive RA, especially long sleep duration. Furthermore, a cohort study from the Taiwan National Health Insurance Research Database suggested that obstructive sleep apnea, a sleep disorder linked to snoring, may increase RA risk, although treating OSA could mitigate this risk [35]. However, we did not find a significant association between snoring and RA risk, which may be due to the self-reported nature of snoring data and the lack of detailed information such as snoring frequency and duration.
The PSS constructed in our study is essential for assessing individualsβ sleep patterns. Sleep behaviors are interrelated and can influence one another, with the body holistically regulating sleep. Changes in one sleep factor often led to compensatory changes in other [36,37]. For example, insomnia could result in shorter sleep duration and increased daytime drowsiness. The PSS developed here provides a comprehensive measure of sleep behavior, capturing the combined impact of five poor sleep factors on RA risk. This score offers a useful framework for assessing overall sleep quality, allowing for more accurate identification of high-risk individuals. Moreover, the simplicity of the scoring algorithm enhances the clarity and applicability of epidemiological findings, increasing their information value [36].Recent studies investigating the relationship between RA and sleep, such as that by Liu et al. [38], have provided valuable insights. Our study contributes to this field by introducing several innovative approaches. First, we included a sensitivity analysis focused on seropositive RA, enriching our understanding of the impact on different RA subtypes. Second, we performed a mediation analysis, which demonstrated that the mGPS acts as a mediator between sleep behaviors and RA risk. Third, we conducted a PSM case-control study to minimize confounding, thus strengthening the reliability of our results. Finally, we employed LDSC to explore genetic correlations between sleep behaviors and RA, uncovering potential shared genetic mechanisms. These methodological innovations help further elucidate the complex interactions between sleep patterns, genetic predisposition, and RA.
Several mechanisms may link sleep disturbances to RA risk, including chronic inflammation. We found that the mGPS, an inflammation-based score, mediated the relationship between sleep and RA risk, although the effect was modest. Sleep and immune system disorders are linked in both directions. Sleep disturbances and immune system disorders are bidirectional: a disrupted immune system can impair sleep, while poor sleep, in turn, can affect immune function [39]. Previous studies have suggested that insomnia and other non-apnea sleep disorders may increase the risk of autoimmune diseases, including RA, systemic lupus erythematosus, and ankylosing spondylitis [17]. Chronic low-grade inflammation associated with poor sleep is thought to contribute to RA risk [40]. Immunological studies indicate that sleep deprivation can elevate the number of circulating T cells and B cells, while disrupting the rhythm of CD4+ regulatory T cells (Tregs) [41,42]. Activated T cells can stimulate osteoclasts, leading to joint destruction. Additionally, the interaction between T cells and B cells can perpetuate the degradation of the synovial membrane [43]. Furthermore, poor sleep can upregulate inflammatory markers such as tumor necrosis factor-alpha (TNF-Ξ±), interleukin (IL)-1, IL-6, IL-8, and IL-17 [44,45]. The finding that women generally exhibit higher levels of inflammatory markers aligns with our observation that women with poor sleep quality may be at greater risk for RA [46]. Dysregulation of IL-6 and TNF-Ξ± is known to disrupt the balance between cartilage and bone matrix degradation in joints [47]. Notably, an animal study confirmed that IL-17 levels remained elevated in sleep-deprived rats even after 7 days of recovery [48]. IL-17 not only triggers an inflammatory response but also enhances RANKL sensitivity by modulating RANK expression, which increases osteoclast activity and bone resorption [49].
Strength and limitations
This study has several strengths. First, we assessed the relationship between sleep and various types of RA by integrating prospective cohort and PSM case-control studies. Second, we developed a sleep score to evaluate the association between overall sleep patterns and RA risk, incorporating multiple sleep behaviors. Third, we explored the combined effects of sleep patterns and genetic risk on RA. Additionally, we identified significant genetic correlations between sleep phenotypes and different RA subtypes. However, our study also has limitations. First, it was conducted among European Caucasians, limiting the generalizability of the findings to other racial groups. Second, sleep patterns were assessed only at baseline, and the use of a single baseline measure to create the sleep score does not account for changes in sleep behaviors over time. Future studies should incorporate multiple time points to better understand how changes in sleep behavior affect RA risk. Third, while we adjusted for known confounders, there may still be unmeasured or unknown confounding factors. Finally, the sleep data were self-reported, introducing potential measurement errors.
Supporting information
Acknowledgments
We thank previous studies for making the summary association statistics data publicly available and we are grateful to all the investigators and participants in UK Biobank for the public individual-level data sets and their contribution to those studies. UK Biobank approval was given for this project (ID: 76875) and can be downloaded from (https://www.ukbiobank.ac.uk/β or https://biota.osc.ox.ac.uk/β).
Data Availability
All relevant data are within the paper and its Supporting Information files. GWAS Summary data on sleep phenotypes and rheumatoid arthritis can be obtained from the corresponding articles, and we list the DOI numbers of the involved articles: sleep duration (https://doi.org/10.1038/s41467-019-08917-4β); difficulty getting up (https://doi.org/10.1038/s41588-018-0333-3β); chronotype (https://doi.org/10.1038/s41467-018-08259-7β); nap during day (https://doi.org/10.1038/s41467-020-20585-3β); insomnia (https://doi.org/10.1038/s41588-022-01124-wβ); snoring (https://doi.org/10.1038/s41467-020-14625-1β); daytime dozing (https://doi.org/10.1038/s41467-019-11456-7β); rheumatoid arthritis (https://doi.org/10.1038/nature12873β). https://doi.org/10.1038/s41467-019-11456-7β The UK Biobank genotype and phenotype data is available on application to the UK Biobank project (ID: 76875) at http://www.ukbiobank.ac.ukβ. Code is available at https://github.com/Cyanyyq/sleep_RAβ.
Funding Statement
The study was supported by Natural Science Foundation of China (82373587, 82173529, 82173598, and 82103922) and a Project of the Priority Academic Program Development of Jiangsu Higher Education Institutions. The funder participated in the study design, data collection and analysis, decision to publish and manuscript writing aspects.
References
Associated Data
Supplementary Materials
Data Availability Statement
All relevant data are within the paper and its Supporting Information files. GWAS Summary data on sleep phenotypes and rheumatoid arthritis can be obtained from the corresponding articles, and we list the DOI numbers of the involved articles: sleep duration (https://doi.org/10.1038/s41467-019-08917-4β); difficulty getting up (https://doi.org/10.1038/s41588-018-0333-3β); chronotype (https://doi.org/10.1038/s41467-018-08259-7β); nap during day (https://doi.org/10.1038/s41467-020-20585-3β); insomnia (https://doi.org/10.1038/s41588-022-01124-wβ); snoring (https://doi.org/10.1038/s41467-020-14625-1β); daytime dozing (https://doi.org/10.1038/s41467-019-11456-7β); rheumatoid arthritis (https://doi.org/10.1038/nature12873β). https://doi.org/10.1038/s41467-019-11456-7β The UK Biobank genotype and phenotype data is available on application to the UK Biobank project (ID: 76875) at http://www.ukbiobank.ac.ukβ. Code is available at https://github.com/Cyanyyq/sleep_RAβ.