Arthritis, Sleep Health, and Systemic Inflammation in Older Men

Abstract

Objective:

This study examined the associations of prevalent arthritis with systemic inflammation in older men, and tested whether sleep health mediates the associations.

Methods:

Cross-sectional data came from 2,562 community-dwelling older men (all were 65 or older; M_age=76yrs) in the Osteoporotic Fractures in Men Study who participated in a sleep ancillary study in 2003–2005. Participants were classified as having osteoarthritis (OA; 24%) or rheumatoid arthritis (RA; 0.7%) based upon self-reported diagnoses and medication use. We constructed a composite score of multidimensional sleep health (i.e., perceived sleep quality, sleepiness, frequency of daytime napping, wake-after-sleep-onset, and sleep duration) measured by both self-report and actigraphy. We also created binary indicators of elevated inflammation using C-reactive protein (CRP; >3mg/L) and interleukin-6 (IL-6; >1.08pg/mL). Analyses controlled for age, diagnosed sleep disorders, BMI, smoking status, relevant medication use, and comorbidities.

Results:

Older men with OA did not have higher risk of elevated CRP or IL-6. However, indirect associations of OA through sleep health were found; OA was associated with poorer sleep health, which was further associated with 16% higher odds of elevated CRP (p<.001) and 12% higher odds of elevated IL-6 (p<.01), after controlling for OA. Older men with RA had higher odds of elevated CRP and IL-6, but the associations were not mediated by sleep health.

Conclusion:

Findings suggest that promoting sleep health may help reduce the risk of systemic inflammation in older men with osteoarthritis.

Introduction

Systemic inflammation is a hallmark of many chronic diseases, such as arthritis (1,2). Arthritis involves painful and disabled joints and is prevalent among older adults (3). Studies report systemic inflammation associated with arthritis and, especially with rheumatoid arthritis (4–6). However, less is known about potential mechanisms linking arthritis and systemic inflammation in older adults who may be particularly vulnerable to systemic inflammation (7). Given that individuals who have arthritis tend to sleep poorly (8,9) and poor sleep is associated with greater risk of systemic inflammation even in healthy young and middle-aged adults (10), sleep health may be an important biobehavioral mechanism linking arthritis and systemic inflammation in older adults.

Sleep health is a multidimensional construct encompassing quantitative and qualitative aspects. This is particularly important when examining older adults’ sleep, due to age-related changes in sleep characteristics (e.g., decreased sleep duration and efficiency) (11). Yet, most studies have focused on testing the effects of a sleep disorder or a single sleep deficiency (e.g. short sleep duration) on health (12–16), limiting the ability to understand how multiple aspects of sleep contribute to health. Emerging literature on sleep suggests that sleep health needs to be understood in multiple dimensions, such as perceived sleep quality, sleepiness, frequency of daytime napping, amount of wakes after sleep onset, and nighttime sleep duration (17,18). For example, an older adult who sleeps 7 hours or more per night may still have poor sleep health if he/she reports low sleep quality and excessive sleepiness, naps during daytimes, and/or wakes frequently after sleep onset. By assessing multiple dimensions of sleep via self-report and actigraphy, the present study examines how older adults’ sleep health plays a role in the link between arthritis and systemic inflammation, measured by C-reactive protein (CRP) and interleukin-6 (IL-6). Higher levels of circulating CRP are regarded as a valid biomarker of systemic inflammation (19), and IL-6 is a key player in systemic inflammation and joint destruction (4).

Elevated CRP and IL-6 may be associated with older adults’ chronic arthritis conditions that may trigger poor sleep health. Two forms of arthritis are most prevalent in older adults—osteoarthritis (OA), a degenerative joint condition and rheumatoid arthritis (RA), an autoimmune condition. OA and RA are each characterized by joint destruction. In both diseases, inflammation is implicated; yet, RA is generally associated with more inflammation compared with OA (20,21). OA by itself is not a systemic inflammatory disease, and studies investigating the relation between OA and systemic markers of inflammation yield conflicting results (22,23). Subjects with OA typically have other chronic diseases and experience associated pain and sleep problems that may increase the risk of systemic inflammation (24,25). Moreover, clinical evidence shows that older adults with OA have the greatest joint pain at the end of the day and such pain may lead to poor sleep at night (26). Thus, poor sleep may be a potential mechanism linking OA and systemic inflammation. In contrast, subjects with RA usually have the greatest pain in the morning or upon waking (27,28). This suggests that the link between RA and systemic inflammation may not be necessarily through poor sleep. However, some studies report poor sleep in patients with RA (8,29,30), thus the relations between RA, sleep health, and systemic inflammation warrant more exploration.

The purpose of the present study was to test the mediating role of sleep health in the link between arthritis (OA and RA) and systemic inflammation (CRP and IL-6) in a sample of community-dwelling older men. Based on the qualitative differences between OA and RA (20), we had different hypotheses regarding the mechanisms of OA and RA leading to systemic inflammation. We hypothesized that OA would be indirectly associated with older men’s higher risks of elevated CRP and IL-6, mediated by poorer sleep health (Hypothesis 1). We also hypothesized that RA would be directly associated with older men’s higher risks of elevated CRP and IL-6 (Hypothesis 2a). We further tested whether the associations would be mediated by poorer sleep health (Hypothesis 2b).

Methods

Participants

Data came from the Osteoporotic Fractures in Men Study (MrOS, http://mrosdata.sfcc-cpmc.net), a prospective cohort study of 5,994 men in the US who were recruited from 2000 to 2002 at 6 geographically dispersed clinical sites (Birmingham, Alabama; Minneapolis, Minnesota; Palo Alto, California; the Monongahela Valley near Pittsburgh, Pennsylvania; Portland, Oregon; and San Diego, California) (31,32). To be eligible, men had to be age ≥ 65 years, were able to walk without assistance, and must not have had a bilateral hip replacement. We used a sub-sample from an ancillary Sleep Study, which was conducted at the six clinical centers between December 2003 and March 2005, and included 3135 participants (33). For the Sleep Study, men were excluded if they regularly used positive pressure or oral appliances during sleep for treatment of sleep apnea, or used overnight nocturnal oxygen therapy (n = 150). Other reasons for non-participation were: death between baseline and the sleep visit (n = 349), terminated study participation (n= 39), declined sleep study (n= 1997), or because the MrOS Sleep Study recruitment goals had already been met prior to enrollment in the Sleep Study and eligible subjects were not invited to participate in the sub-study (n = 324) (33). Inflammatory marker assays were measured from serum collected at the Sleep Visit in 2562 men, which comprised the final analytic sample of this study.

Procedure

Participants completed questionnaires providing information on sociodemographics, chronic diseases, smoking status, perceived sleep quality and sleepiness, and medication use. Participants were asked to wear actigraphs on the non-dominant wrist for a minimum of 5 consecutive 24-hour periods; devices were removed only for bathing or during water sports. ActionW-2 software (Ambulatory Monitoring, Inc., Ardsley, NY) was used to score actigraphy data, and details of the scoring algorithms used have been published elsewhere (34). Inter-scorer reliability for the scoring of these data has been previously found to be high in our sample (intraclass coefficient = 0.95) (34). Blood was collected during morning clinic visits, after an overnight fast, and serum samples were frozen at −80°C until assays were performed. The institutional review boards at each clinic site approved the study, and written informed consent was obtained from all participants.

Measures

Arthritis (predictors).

We asked participants whether they ever had osteoarthritis (OA) or rheumatoid arthritis (RA). Responses were coded as yes (=1) or no (=0). For RA, 205 men responded that they ever had the condition. Considering that many older adults believe that they have RA even when they actually do not (35), we verified their medication list. Out of 205 men, only 18 men took medications specific to RA and thus were considered as having RA in our analyses. Those medications used to determine RA were prednisone and oral disease remissive agents or Disease-modifying antirheumatic drugs (DMARDs) include Methotrexate, Hydroxychloroquine, Plaquenil, leflunomide, and/or biologic anti-cytokine therapies including anti-tumor necrosis factor antibodies, and anti-interleukin 6 antibodies. Participants were asked to bring all current medications they used within the past month to the clinic during the Sleep Visit. Each medication was matched to its ingredient(s) based on the Iowa Drug Information Service (IDIS) Drug Vocabulary (College of Pharmacy, University of Iowa, Iowa City, IA).

Systemic inflammatory markers (outcomes).

All assays were performed at the Johns Hopkins University Clinical Research Unit (CRU) Core Laboratory. C-reactive protein (CRP) was measured using enzyme linked immunosorbent assays (ELISAs) from ALPCO (Salem, NH). This assay utilizes a sandwich enzyme immunoassay, in which plate wells are coated with polyclonal antibodies to CRP. The inter-assay CVs ranged from 11.6 to 13.8%. IL-6 was assayed using the Human ProInflammatory I 4-Plex Ultra-Sensitive Kit by Meso Diagnostics (MSD; Rockville MD; catalog #K15009C-4). The sensitivity of the assay was 0.22 pg/mL and the inter-assay CVs ranged from 2.0 to 9.9%. We created binary indicators of elevated risk of CRP (>3mg/L, following a clinical cutoff that has been related to higher risk of CVD (36); 26% of the sample) and IL-6 (>1.08pg/mL; calculated from the sample median (33)).

Sleep health (mediators).

Five key dimensions of sleep health were assessed via self-report and actigraphy following previously suggested criteria (17,18). First, self-reported sleep quality was measured by the Pittsburgh Sleep Quality Index (PSQI, Range = 0–21) (37). Second, self-reported sleepiness was measured by the Epworth Sleepiness Scale (ESS, Range = 0–24) (38). Third, actigraphy-assessed daytime napping was measured as weekly frequency. A nap was defined as out-of-bed sleep for 5 consecutive minutes or more. Fourth, actigraphy-assessed wake-after-sleep-onset (WASO) was measured in minutes per night. Following previous studies (39,40), we used total amount of wakes in bed (after sleep onset) per night and calculated a weekly average for each participant. Lastly, actigraphy-assessed nighttime sleep duration was measured in minutes per night. Using previously used standard cut-off points (33), we created binary indicators of poor sleep quality (PSQI >5), excessive sleepiness (ESS > 10), frequent daytime napping (≥ 2 times/week), long wake-after-sleep-onset (≥ 90minutes/night), and short nighttime sleep duration (≤ 6 hours/night). We summed across these binary indicators to construct a composite score of sleep health, such that higher scores represented poorer sleep health (Range = 0–5).

Covariates.

We controlled for age (in years; centered at the sample mean), any diagnosed sleep disorders (1=yes, 0=no), body mass index (BMI; kg/m²), current smoking status (1=yes, 0=no), and medication use. Given that the majority were Caucasian (over 91%), we did not control for race/ethnicity. Medications that were previously considered to potentially affect levels of inflammatory markers and sleep (26) were entered as covariates, including: antidepressants, benzodiazepines, sedatives/hypnotics, other medications used for sleep, nonsteroidal anti-inflammatory drugs (NSAIDs), and corticosteroids. We included binary indicators of each medication (1=using, 0=not using). In addition, we took into account co-morbidities defined as having two or more chronic diseases in 19 conditions including OA, RA, osteoporosis, prostatitis, asthma, diabetes, chronic bronchitis, COPD/emphysema, glaucoma, high or low thyroid, liver disease, Parkinson’s disease, kidney disease, hypertension, congestive heart failure, heart attack, stroke, angina or chest pain (1=yes, 0=no). For example, in our sample, 4 participants had both OA and RA; they were coded as having co-morbidities.

Statistical Analyses

We used SAS PROCESS macro that can test the indirect association of X (predictor; arthritis) with Y (outcome; systemic inflammation) mediated by M (mediator; sleep health) based on the bootstrapping method (41). This method shares similarities to more traditional causal step mediation approaches, such as Baron and Kenny’s approach, but it has several advantages. First, it allows for simultaneous testing between X and Y, X and M, and M and Y in a single analytic step, which can inform a more comprehensive understanding of the relations between the variables. The traditional causal step approach is based on individual hypothesis tests involving X→M (“a”), M→Y (“b”), and X→Y (“c”), where the ability to claim M as a mediator is contingent on the significance of the three paths. Unlike the traditional mediation approach, our method allows for the estimation of the indirect association, based on “a×b”, not based on individual hypothesis testing, and even without the total association of X with Y (42,43). Moreover, the bootstrapping method produces a bias-corrected confidence interval for the indirect association. The bootstrapping method is especially useful when the true distribution of a behavior is not known, as it generates an empirically derived representation of the sampling distribution by repeated resampling based on the original sample; this empirical representation is used for the construction of a robust confidence interval (see (41) for more details). In all models, we set the number of bootstrap samples to 10,000. In result tables, B indicates beta coefficients predicting the continuous variable of sleep health. Exp(B) indicates exponentiated betas (odds ratios) predicting binary indicators of elevated systemic inflammation.

Results

Table 1 shows descriptive statistics by arthritis conditions. In our sample of 2562 men, 24% of participants reported that they ever had osteoarthritis (OA); among 8% of participants (n=205) who reported they ever had rheumatoid arthritis (RA), 0.7% (n=18) were actually on RA-specific medications. These numbers were similar to the national prevalence rates. (44,45) The mean age was 76 years (SD = 5.5) with no difference by arthritis status. Participants with OA were more likely to have diagnosed sleep disorders (p < .001), use antidepressants (p < .001), benzodiazepines (p < .001), sleep medications (p < .01), NSAIDs (p < .001) and corticosteroids (p < .01), and have co-morbidities (p < .001) compared to without OA. Those with RA were more likely to use corticosteroids (p < .001) compared to those without RA. All these demographic and medical characteristics were included as covariates in subsequent models. The percentage of older men who had elevated CRP and IL-6 did not differ by OA status, but differed by RA status such that those with RA were more likely to exhibit both elevated CRP and IL-6 than those without RA. In terms of sleep health dimensions, participants with OA were more likely to report poor sleep quality (p < .001) and excessive sleepiness (p < .05) than those without OA; participants with RA were more likely to report excessive sleepiness (p < .01) than those without RA. Combining these multiple dimensions of sleep health, participants with OA (but not RA) exhibited poorer sleep health composite scores (p < .001) than those without OA. Results were similar when we compared OA or RA subjects to those who had neither OA nor RA (n=1924, 75%), except for one difference: more RA subjects used NSAIDs (39%) than arthritis-free subjects (15%, p < .01).

Table 1.

Descriptive statistics by history of arthritis

	Osteoarthritis					Rheumatoid Arthritis
	Ever had (24.3%)		Never had (75.7%)		Diff Test4	Ever had (0.7%)		Never had (99.3%)		Diff Test4
	M or %	(SD)	M or %	(SD)		M or %	(SD)	M or %	(SD)
Covariates
Age (years)	76.19	(5.49)	76.41	(5.51)	ns	78.83	(6.48)	76.34	(5.49)	ns
Have any diagnosed sleep disorders (Yes)	12%		8%		***	6%		9%		ns
BMI (kg/m2)	27.39	(3.83)	27.08	(3.74)	ns	26.01	(2.55)	27.16	(3.77)	ns
Smoking status (Yes)	1%		2%		ns	0%		2%		ns
Antidepressants (Using)	12%		7%		***	17%		8%		ns
Benzodiazepines(Using)	8%		4%		***	6%		5%		ns
Sedatives/hypnotics (Using)	3%		2%		ns	6%		2%		ns
Sleep medications (Using)	15%		11%		**	6%		12%		ns
Nonsteroidal anti-inflammatory drugs (Using)	40%		15%		***	39%		21%		ns
Corticosteroids (Using)	12%		8%		**	78%		9%		***
Co-morbidities3 (Yes vs. No)	37%		12%		***	11%		18%		ns
Systemic Inflammation
Elevated CRP (> 3mg/L)	25%		26%		ns	67%		26%		***
Elevated IL-6 (>1.08pg/mL)	51%		50%		ns	89%		50%		***
Sleep Health
Poor sleep quality (PSQI1 > 5)	53%		41%		***	33%		44%		ns
Excessive sleepiness (ESS2 > 10)	15%		11%		*	33%		12%		**
Actigraphy long WASO (≥ 90 minutes/night)	34%		31%		ns	22%		31%		ns
Actigraphy short sleep duration (≤ 6 hours/night)	34%		30%		ns	39%		31%		ns
Actigraphy frequent napping (≥ 2 times/week)	47%		47%		ns	56%		47%		ns
Composite sleep health score (0 to 5=poorer)	1.83	(1.22)	1.60	(1.17)	***	1.83	(1.29)	1.66	(1.19)	ns

Note. N=2562 older men.

¹PSQI=Pittsburgh Sleep Quality Index.

²ESS=Epworth Sleepiness Scale.

³Co-morbidities were defined as having two or more diseases in 19 conditions including osteoarthritis and rheumatoid arthritis.

⁴Diff Test: Difference tests compared those who ever had each arthritis condition (either osteoarthritis or rheumatoid arthritis) to those without the condition.

ns p ≥ .05,

^*p < .05,

^**p < .01,

^***p < .001.

Table 2 presents results from mediation models examining the associations of OA with elevated CRP and IL-6 through sleep health. Older men with OA exhibited poorer sleep health than those without OA (p < .05), even after controlling for age, having diagnosed sleep disorders, BMI, smoking, using antidepressants and sleep medications. Poorer sleep health was further associated with 16% higher odds of elevated CRP (p < .001), after controlling for OA and independent of BMI, smoking, using benzodiazepines corticosteroids and sleep medications. Poorer sleep health was also associated with 12% higher risk of elevated IL-6 (p < .01), after controlling for OA and covariates. Older age, higher BMI, smoking, and using corticosteroids each significantly predicted higher risk of elevated IL-6. Before including sleep health in the model, there were no significant associations of OA with either CRP or IL-6 (Figure 1). However, when we considered sleep health, the whole model revealed significant indirect associations of OA on higher risks of elevated CRP and IL-6 through sleep health: Indirect Association predicting CRP = 0.0202, SE = 0.0104, 95% CI = [0.0048, 0.0471] and Indirect Association predicting IL-6 = 0.0153, SE = 0.0086, 95% CI = [0.0031, 0.0380], respectively. Thus, hypothesis 1 was supported.

Table 2.

Results of the cross-sectional mediation model examining the mediating role of sleep health on the link between the history of osteoarthritis and the risks of elevated C-reactive protein (CRP) and Interleukin-6 (IL-6)

	M: Sleep Health (0 to 5=poorer)			Y1: Risk of Elevated CRP			Y2: Risk of Elevated IL-6
	B		95% CI	Exp(B)		95% CI	Exp(B)		95% CI
Intercept	−0.17	***	[−0.23, −0.11]	0.33	***	[0.29, 0.37]	0.96		[0.86, 1.07]
X: Osteoarthritis (Having OA vs. Never had OA)	0.14	*	[0.03, 0.25]	0.83		[0.66, 1.04]	1.02		[0.83, 1.25]
M: Sleep Health (0 to 5=poorer)	--		--	1.16	***	[1.07, 1.25]	1.12	**	[1.04, 1.20]
Age (in years)	0.02	***	[0.02, 0.03]	1.01		[1.00, 1.03]	1.08	***	[1.06, 1.10]
Have any diagnosed sleep disorders (Yes vs. No)	0.40	***	[0.24, 0.57]	0.99		[0.72, 1.37]	1.01		[0.75, 1.35]
BMI (kg/m2)	0.07	***	[0.05, 0.08]	1.08	***	[1.05, 1.11]	1.11	***	[1.08, 1.13]
Smoking status (Yes vs. No)	0.49	**	[0.17, 0.81]	2.66	***	[1.49, 4.73]	2.56	**	[1.4, 4.69]
Antidepressants (Using vs. Not using)	0.26	**	[0.09, 0.43]	1.20		[0.85, 1.67]	1.04		[0.76, 1.42]
Benzodiazepines (Using vs. Not using)	0.11		[−0.12, 0.34]	2.05	**	[1.31, 3.20]	1.12		[0.73, 1.71]
Sedatives/hypnotics (Using vs. Not using)	0.20		[−0.14, 0.53]	0.91		[0.44, 1.91]	0.74		[0.4, 1.38]
Sleep medications (Using vs. Not using)	0.27	**	[0.11, 0.44]	0.66	*	[0.47, 0.95]	0.93		[0.69, 1.27]
NSAIDs1 (Using vs. Not using)	0.10	†	[−0.01, 0.21]	0.99		[0.79, 1.25]	0.83	†	[0.67, 1.01]
Corticosteroids (Using vs. Not using)	0.11		[−0.04, 0.26]	1.54	**	[1.15, 2.07]	1.93	***	[1.45, 2.59]
Co-morbidities2 (Yes vs. No)	0.003		[−0.12, 0.12]	1.16		[0.91, 1.48]	1.05		[0.84, 1.30]
	R2 = 0.09			-2LL = 2834.58			-2LL = 3344.10
	F= 20.97***			Model LL = 100.31			Model LL = 204.81

Note. N=2562 older men; 2560 observations were used due to missing values in the variables.

X: predictor, M: mediator, Y: outcome.

¹Nonsteroidal anti-inflammatory drugs.

²Co-morbidities were defined by having two or more diseases in 19 conditions including osteoarthritis and rheumatoid arthritis.

^†p < .10,

^*p < .05,

^**p < .01,

^***p < .001.

Figure 1. Pathways linking the associations of osteoarthritis with elevated systemic inflammation through sleep health.

Note. N=2562 older men. CRP: C-reactive protein. IL-6: Interleukin-6. CRP and IL-6 were modeled separately. Analyses adjusted for covariates. Dotted lines indicate non-significant paths. * p < .05, ** p < .01, *** p < .001.

Table 3 presents results from mediation models examining the associations of RA with elevated CRP and IL-6 mediated by sleep health. Older men with RA did not exhibit poorer sleep health than those without RA (first column). However, poorer sleep health was associated with 15% higher risk of elevated CRP (p < .001) and 12% higher risk of elevated IL-6 (p < .01). Before including sleep health in the model, there were significant direct associations of RA with CRP (p < .01) and IL-6 (p < .05), but the associations were not mediated by sleep health (Figure 2). Thus, hypothesis 2a was supported, but hypothesis 2b was not supported.

Table 3.

Results of the cross-sectional mediation model examining the mediating role of sleep health on the link between the history of rheumatoid arthritis and the risks of elevated C-reactive protein (CRP) and Interleukin-6 (IL-6)

	M: Sleep Health (0 to 5=poorer)			Y1: Risk of Elevated CRP			Y2: Risk of Elevated IL-6
	B		95% CI	Exp(B)		95% CI	Exp(B)		95% CI
Intercept	−0.16	***	[−0.21, −0.10]	0.32	***	[0.28, 0.36]	0.96		[0.86, 1.07]
X: Rheumatoid Arthritis (Having RA vs. Never had RA)	0.10		[−0.44, 0.64]	5.10	**	[1.84, 14.03]	5.81	*	[1.29, 26.12]
M: Sleep Health (0 to 5=poorer)	--		--	1.15	***	[1.07, 1.25]	1.12	**	[1.04, 1.20]
Age (in years)	0.02	***	[0.02, 0.03]	1.01		[0.99, 1.03]	1.08	***	[1.06, 1.10]
Have any diagnosed sleep disorders (Yes vs. No)	0.41	***	[0.25, 0.57]	0.99		[0.72, 1.38]	1.01		[0.75, 1.36]
BMI (kg/m2)	0.07	***	[0.05, 0.08]	1.08	***	[1.06, 1.11]	1.11	***	[1.08, 1.13]
Smoking status (Yes vs. No)	0.49	**	[0.17, 0.81]	2.69	***	[1.51, 4.79]	2.56	**	[1.40, 4.71]
Antidepressants (Using vs. Not using)	0.27	**	[0.10, 0.44]	1.16		[0.83, 1.63]	1.04		[0.76, 1.42]
Benzodiazepines (Using vs. Not using)	0.13		[−0.11, 0.36]	1.99	**	[1.28, 3.11]	1.12		[0.73, 1.71]
Sedatives/hypnotics (Using vs. Not using)	0.21		[−0.13, 0.54]	0.87		[0.41, 1.82]	0.73		[0.39, 1.36]
Sleep medications (Using vs. Not using)	0.27	**	[0.10, 0.43]	0.69	*	[0.48, 0.98]	0.95		[0.70, 1.29]
NSAIDs1 (Using vs. Not using)	0.14	*	[0.03, 0.25]	0.93		[0.75, 1.16]	0.82	†	[0.67, 1.00]
Corticosteroids (Using vs. Not using)	0.11		[−0.05, 0.27]	1.38	*	[1.02, 1.86]	1.80	***	[1.35, 2.43]
Co-morbidities2 (Yes vs. No)	0.04		[−0.07, 0.16]	1.11		[0.88, 1.40]	1.06		[0.86, 1.31]
	R2 = 0.09			-2LL = 2827.24			-2LL = 3337.99
	F= 20.41***			Model LL = 108.25			Model LL = 212.31

Note. N=2562 older men; 2561 and 2560 observations were used in the models predicting CRP and IL-6, respectively, due to missing values in the variables.

X: predictor, M: mediator, Y: outcome.

¹Nonsteroidal anti-inflammatory drugs.

²Co-morbidities were defined by having two or more diseases in 19 conditions including osteoarthritis and rheumatoid arthritis.

^†p < .10,

^*p < .05,

^**p < .01,

^***p < .001.

Figure 2. Pathways linking the associations of rheumatoid arthritis with elevated systemic inflammation through sleep health.

Note. N=2562 older men. CRP: C-reactive protein. IL-6: Interleukin-6. CRP and IL-6 were modeled separately. Analyses adjusted for covariates. Dotted lines indicate non-significant paths. * p < .05, ** p < .01, *** p < .001.

We conducted supplementary analyses using continuous measures of CRP and IL-6. Results were consistent with those from binary models. Older men with OA had poorer sleep health (B = 0.14, SE = 0.06, p < .05); in turn, poorer sleep health was associated with higher levels of CRP (B = 0.33, SE = 0.09, p < .001) and IL-6 (B = 0.38, SE = 0.12, p < .01). Older men with RA exhibited higher levels of CRP (B = 2.71, SE = 1.27, p < .05) and IL-6 (B = 5.04, SE = 1.64, p < .01), but these associations were not mediated by sleep health. These associations also held after removing potential outliers of CRP (> 10mg/L; clinical cut-point) and IL-6 (> 4.3pg/mL; 5^th percentile). In addition, for the RA-inflammation association, we repeated our analyses with a larger sample of 205 men who reported they have ever had RA. The results were consistent with those based on a smaller sample (18 RA men). Older men who reported they ever had RA had higher odds of elevated CRP and IL-6 than those who never had RA, but the associations were not mediated by sleep health.

Discussion

In a cohort of community-dwelling older men, we found that having osteoarthritis (OA) was indirectly associated with higher risk of systemic inflammation through poorer sleep health. Conversely, rheumatoid arthritis (RA) was directly associated with higher risks of systemic inflammatory markers (CRP and IL-6), but the associations were not mediated by poorer sleep health. Although both OA and RA are chronic diseases characterized by joint destruction and inflammation, qualitative differences between them (20) might have resulted in the different pathways leading to systemic inflammation.

Our results showed that individuals with OA did not exhibit a higher risk of systemic inflammation before we statistically accounted for their sleep health. However, individuals with OA did have poorer sleep health, and poorer sleep health in OA subjects was associated with increased risks of elevated CRP and IL-6. Previous studies examining the relation between OA and systemic markers of inflammation have reported inconsistent results where some found significant associations and others did not (22,23). The current study reveals the indirect pathway from OA to systemic inflammation through poorer sleep health, which may explain the inconsistency in previous results. Previous studies also have observed sleep disturbances in subjects with OA (24,25,46); the current study extends this line of work with a new finding that poor sleep health in OA subjects is further associated with elevated risk of systemic inflammation. Given that OA is more likely to involve joint pain at the end of the day (26), perhaps more nighttime pain contributed to poorer sleep health in those with OA (9,47). It may also be that the link between OA and systemic inflammation depends on the severity of and sensitivity to pain (9,23,48). We did not assess pain in this study, however; future studies could examine how specific dimensions of OA pain (i.e., sensitivity, severity, and time of pain) may be associated with sleep health and systemic inflammation.

This study found significant direct associations of RA with higher risks of elevated CRP and IL-6. The relationship between RA and inflammation is not new (4). However, there is relatively little research examining the associations of RA and inflammatory markers in a sample of community-dwelling older men who may be particularly vulnerable to inflammation. In our sample, even among those without a history of RA or OA, half had elevated IL-6 (Table 1). We further tested whether these associations were mediated by sleep health and they were not. Although some studies have found RA to be correlated with poor sleep (8,29,30), this correlation may be confounded by many factors including fatigue, physical inactivity, obesity, and depression, all of which are often observed in RA subjects (28,29). When controlling for some of these confounding factors (e.g., BMI, an extensive list of medications) in the current analyses, we found there was no direct association between RA and sleep health. We need to be cautious though in interpreting this null association between RA and sleep health, as our group size for RA was small (n=18) and thus we might have been underpowered to detect this association.

Our findings contribute to the literature on arthritis and treatment by showing that older men with a history of OA may have a higher risk of systemic inflammation if they have poorer sleep health. As there is a positive feedback loop between inflammation and joint pain (49), we could expect that promoting sleep health may also have a positive influence on their level of daily pain. At the clinical level, healthcare professionals could include a brief sleep assessment to identify individuals who have poor sleep health, a significant risk factor for systemic inflammation found in this study. More research is needed to replicate these findings, as well as trials to systematically intervene on sleep to potentially augment current arthritis-related pain treatment regimens. Further strengths of this study include a relatively large sample of community-dwelling older men who provided serum samples of inflammatory markers, and the use of objective and subjective sleep health measures to test whether the arthritis–systemic inflammation association is mediated by multidimensional sleep characteristics. Controlling for a wide array of relevant medications and co-morbidities also increases our confidence in interpreting these results. Overall, our findings suggest that promoting sleep health may have benefits in individuals with OA.

Limitations of this study include relying on self-reported history of OA and RA without information on clinical diagnoses at the time of this study. A close examination of medication history for those who reported having RA would lower this concern. Future research that verifies the presence of OA and RA via diagnostic measures may yield more clinically meaningful results. The prevalence of OA (24%) and RA (0.7%) in this study sample was similar to the national statistics (44,45); however, a small group size of RA may lack statistical power to detect the relationship and be prone to type II error. Thus, these results may underestimate the potential adverse association of RA with sleep health. Future research with a large sample of RA subjects may overcome this issue. An additional limitation is that these data were cross-sectional, and thus we cannot determine temporality between arthritis, sleep health, and systemic inflammation. Although our statistical analyses imply that arthritis is a predictor, sleep health is a mediator, and systemic inflammatory markers are outcome variables, causality can operate in other directions. Lastly, since this study sample consisted of older men who were mostly white (over 91%), our findings may not generalize to younger adults, older women, or other ethnic minority groups.

Conclusion

This study found that men with osteoarthritis have poorer sleep health, and this in turn was associated with higher risk of systemic inflammation. Based on this finding, future intervention studies could test whether promoting sleep health may help reduce the risk of systemic inflammation in older men with osteoarthritis, which is the most common cause of disability in later life (50). Men with rheumatoid arthritis also have a higher risk of systemic inflammation, but this association was not through poorer sleep health. Future research is needed to explore other potential mechanisms linking rheumatoid arthritis and systemic inflammation.

Appendix.

Levels of C-reactive protein (CRP) and Interleukin-6 (IL-6) by osteoarthritis and rheumatoid arthritis conditions

	Total Sample			Osteoarthritis						Rheumatoid Arthritis
				Ever had (24.3%)			Never had (75.7%)			Ever had (0.7%)			Never had (99.3%)
	Mean	Median	(SD)	Mean	Median	(SD)	Mean	Median	(SD)	Mean	Median	(SD)	Mean	Median	(SD)
CRP (mg/L)	2.78	1,52	(5.25)	2.98	1.61	(6.89)	2.72	1.48	(4.60)	5.81	4.39	(4.54)	2.76	1.50	(5.25)
IL-6 (pg/mL)	1.90	1.08	(6.79)	2.57	1.09	(12.52)	1.69	1.08	(3.22)	7.29	3.11	(11.78)	1.86	1.07	(6.73)

Significance and Innovations.

• In a large cohort of community-dwelling older men (age ≥ 65 years), having OA was associated with experiencing poor sleep health, and poor sleep health was further associated with having higher risk of elevated CRP and IL-6.

• Having RA was directly associated with having higher risk of elevated CRP and IL-6; this association was not mediated by poor sleep health after accounting for a wide array of relevant medications and co-morbidities.

• This study reveals different potential mechanisms by which OA and RA may increase risk for systemic inflammation.

• Promoting sleep health in men with OA may help reduce the risk of systemic inflammation.