The relationship between restless sleep and symptoms of the knee: data from the Osteoarthritis Initiative

Abstract

Objective:

To examine the associations between restless sleep and knee symptoms among individuals with radiographically confirmed KOA.

Methods:

Cross-sectional and longitudinal associations were examined using Osteoarthritis Initiative (OAI) data. Participants with radiographic KOA (n=2,517) were asked how often sleep was restless in the past week over the four years and the Western Ontario and McMaster Universities Arthritis Index (WOMAC) was used to measure knee symptoms. Adjusted β coefficients (aβ) and 95% confidence intervals (CI) were derived from generalized estimating equations (GEEs) models stratified by sex.

Results:

One in 7 participants reported ≥3 nights with restless sleep. Cross-sectional analyses indicated that restless sleep 5–7 nights was associated with worse symptoms (Women: pain: aβ: 1.93, 95% CI: 1.12–2.74, stiffness: aβ: 0.57, 95% CI: 0.19–0.94, physical function: aβ 5.68, 95% CI: 3.09–8.27; Men: pain: aβ=1.85, 95% CI: 0.85–2.86; stiffness: aβ: 0.63, 95% CI: 0.15–1.12; physical function: aβ 5.89, 95% CI: 2.68–9.09) compared to <1 night. Longitudinal analyses confirmed that more nights with restless sleep was associated with worse pain (P trend =0.01) and function (P trend =0.04) in women and physical function in men (P trend = 0.04), although estimates did not meet thresholds for minimal clinically meaningful differences.

Conclusion:

While the analysis of cross-sectional data supported the association between restless sleep and KOA symptoms, such relationships were not confirmed in more robust longitudinal analysis. Further research examining whether sleep quality, duration, or disorders is associated with worsening symptoms in persons with KOA is warranted.

Introduction

Knee osteoarthritis (KOA) is a common, disabling musculoskeletal disorder [1, 2]. The incidence rate of symptomatic KOA ranges between 40 to 1,020 per 100,000 person years [3, 4]. KOA affects 37% of American adults (radiographic KOA Kellgren-Lawrence grade ≥2) [5]. As a progressive joint disease, KOA is a leading cause of disability in adults [6]. Economic costs for osteoarthritis such as work loss [7] and healthcare expenses including hospitalizations [8] is known. The human toll of KOA with respect to symptoms such as pain, stiffness, and physical functioning [9, 10] has been documented. However, despite the recent findings [11–13], studies examining the relationship between sleep quality and symptoms in patients with KOA are limited and longitudinal studies are lacking.

Presently, there is no cure for osteoarthritis and there are no Food and Drug Administration approved disease-modifying treatments. For those with KOA, treatment goals focus on improving quality of life by addressing symptoms such as pain, stiffness, and physical functioning [14]. Yet for osteoarthritis clinical care, treatment efficacy demonstrated in clinical trials does not readily translate into “real world” effectiveness [15, 16]. This reality underscores the need to understand the role of other potentially modifiable factors that may be associated with KOA symptoms. Addressing poor sleep quality as a potentially modifiable factor makes sense for several reasons. First, it is a common phenomenon in those with KOA [17]. Second, recent work suggests that reallocating sleep time with moderate physical activity may reduce pain [18]. Third, there are proven interventions to improve sleep quality [19].

We used data from the Osteoarthritis Initiative (OAI), a multi-site longitudinal study which recruited a large number of persons with radiographically confirmed KOA. Self-reported restless sleep was included at each visit, as were validated tools for patient-reported KOA symptoms. This study sought to describe longitudinal changes in restless sleep over time and to determine the extent to which restless sleep was associated with symptoms among people with radiographically confirmed KOA. Using both cross-sectional and longitudinal study designs, we hoped to further understand the bidirectional association between restless sleep and KOA symptoms.

Materials and Methods

Study sample

From 2004–2006, the OAI study enrolled 4,796 participants from Baltimore, MD; Columbus, OH; Pittsburgh, PA; and Pawtucket, RI. The OAI protocol provides details about the cohort study (https://nda.nih.gov/oai/study-details). Briefly, adults aged 45 to 79 years of age with established KOA or who were at high risk for developing KOA were enrolled. Annual follow-ups tracked the development or progression of KOA. We identified participants with radiographically confirmed KOA, which we defined as having a Kellgren-Lawrence grade (K-L) ≥2) at baseline (n=2,548). For the analysis of baseline cross-sectional data, participants with missing information on restless sleep were excluded (n=31). For the analysis of longitudinal data (annual visit from baseline to 48 months), participants with missing restless sleep data (n=572) or symptom measures (n=437) were excluded. The final analytic samples consisted of 2,517 participants for the baseline cross-sectional analyses and 1,882 participants for the longitudinal analyses.

Sleep Measure

Restless sleep was evaluated using a question embedded in the Center for Epidemiologic Studies Depression Scale (CES-D) [20]. The OAI included the annual 20-item CES-D [20] questionnaire from the baseline to the 48-month OAI visit. The participants were asked how often sleep in the past week was restless (rarely/none: <1 night; sometimes: 1–2 nights; much of the time: 3–4 nights; most or every night: 5–7 nights). This item has been previously used as a stand-alone measure of restless sleep [21, 22] and has been shown to be “strongly and independently” associated with sleep symptoms [23].

KOA Symptoms

Pain, stiffness, and physical function were assessed using the WOMAC instrument. With a 5-point Likert scale, the range of scores for the WOMAC measures was 0 to 20 for pain, 0 to 8 for stiffness, and 0 to 68 for physical function [24]. Higher numbers indicated worse symptoms. For the longitudinal analyses, yearly changes in symptoms were assessed from each visit relative to the previous visit (e.g., the change in WOMAC pain score at year 1 was calculated as WOMAC pain score measured at year 1 minus WOMAC pain score measured at baseline). Patient-reported symptoms were collected separately for each knee. We selected an index knee for use in the analyses. If participants had only one radiographically confirmed OA knee at baseline (right knee only 613; left knee only 527), symptom measures for that knee were used. For participants with OA radiographically confirmed in both knees at baseline (n=1,340), we used the knee with greater pain measured by Western Ontario and McMaster Universities Arthritis Index (WOMAC) [24] pain scale (right knee 534; left knee 445). For the participants with radiographically confirmed OA in both knees and equal WOMAC pain scores (n=361), we selected the knee with worse K-L grade (right knee 299, left knee 62).

Covariates

Age, sex, race/ethnicity (i.e., non-Hispanic White, non-Hispanic Black, other), annual household income (i.e., < $25,000, $25,000–50,000, > $50,000), and education (i.e., ≤ high school, some college or college graduate, and ≥graduate school) were measured at baseline and treated as time invariant. Disease severity was measured based on K-L grade and joint space width [25], with implausible values (e.g. the distance between plateau and rim was > 6.5mm) forced to missing. We considered multi-joint involvement present if participants had frequent pain, aching, or stiffness in ≥ two joints other than knee. Prior knee injuries were considered present if participants reported limited ability to walk for ≥ two days. A positive history of knee surgery included arthroscopy, ligament repair or meniscectomy. Evidence of depressive symptoms were recorded if scores for Centers for Epidemiologic Studies Depression Scale were ≥16 [26]. Non-steroidal anti-inflammatory drugs (NSAIDs) use was assessed and recorded for the previous 30 days before the visit. The 12-item Short-Form Health Survey (SF-12) [27] provided summary scores for physical and mental health (range: 0 to 100; higher scores correspond to better health status). Measures of height and weight were used to calculate body mass index (BMI) [weight (kg)/height (m)²; < 25 kg/m², 25 to < 30 kg/m², ≥ 30 kg/m²]. Comorbidity status was measured using Charlson index and categorized into 0, 1, and ≥ 2 [28]. We used the self-reported Physical Activity Scale for the Elderly (PASE) which has 26 items including occupational, household, and leisure activities over the past week [29]. Higher PASE scores indicating greater activities.

Statistical analyses

We first described the sociodemographic and clinical characteristics of the baseline sample stratified by restless sleep. We used descriptive statistics with means and standard deviations (SD) calculated for variables having a continuous distribution and percentages for categorical variables. Because our goal was to describe the sample, rather than to test any specific hypothesis testing, p-values were not calculated. Longitudinal studies that show the extent of switching between restless sleep categories in patients with osteoarthritis are scarce. For this reason, we conducted a descriptive analysis to describe the extent to which participants consistently report the same levels of restless sleep. We also calculated correlations between the covariates of interest to understand (and rule out) the potential for multicollinearity in the modeling phase of the analysis.

To examine associations between restless sleep and KOA symptoms knee osteoarthritis, we first conducted a cross-sectional analysis. These analyses helped put our findings in context with previous research [17, 18] and generated additional hypotheses. We then conducted longitudinal analyses with repeated measures over the four years to further understand the temporal relationship between restless sleep and yearly changes in knee symptoms. We also stratified these analyses by sex because previous research have shown the potential differences in central pain mechanism in patients in inflammatory arthritis as well as the relationship between sleep quality and pain in women and men [30, 31].

For the cross-sectional component of this study, we developed separate models for each of the three symptom variables (i.e., pain, stiffness, physical functioning). For each outcome, we developed a crude linear model and then an adjusted model. The beta coefficients and corresponding 95% confidence intervals (CIs) were derived for each restless sleep category compared to participants reporting rarely or no restless sleep (<1 night in 7 nights preceding the visit). Positive and negative beta coefficients corresponded to higher and lower WOMAC subscale scores in pain, stiffness, and physical function, respectively. While a range of potential confounders were considered, we used an iterative, but not computer-driven approach and included only the subset of confounders that altered the measure of association > 10%.

For longitudinal analyses, we first conducted a descriptive analysis to evaluate the extent to which reports of restless sleep varied over the four visits. Then, we conducted two sets of longitudinal analyses to explore the potential bidirectional relationship between knee symptoms and restless sleep [32, 33]: 1) treated restless sleep as the primary determinant and KOA symptoms as the continuous outcomes; and 2) KOA symptoms were the primary exposure variables and restless sleep as the multinomial outcome. Measurements were reported for up to four assessments (year 1 to year 4) for each individual. Generalized estimating equations (GEE) models were used to examine the associations between both repeated measures of restless sleep (rarely or no restless sleep was the reference group) and yearly changes in knee symptoms [34]. We assumed an exchangeable within-subject correlation structure and tested the robustness of estimates assuming auto-regressive and unstructured correlation structures [34]. All models were adjusted for age, sex, race/ethnicity, K-L grade, body mass index, CES-D, and physical activity scores. Adjusted β coefficients and corresponding 95% confidence intervals (CI) were estimated. For models treating KOA symptoms as outcomes, positive β coefficients corresponded to the absolute increases (worsening) of pain, stiffness, and physical function. Negative β coefficients corresponded to the absolute decreases (improvement) of symptoms. For models treating restless sleep as outcome measures, we fit separate multinomial models using each KOA symptom as the determinants with rarely or no restless sleep treated as the reference group for the outcome variable. From these models, adjusted odds ratios (aOR) and corresponding 95% confidence intervals (CI) were estimated.

Results

Six percent reported having restless sleep most nights (5–7 nights in past week) and 11.4% reported having restless sleep 3–4 nights per week (Table 1). The average age of those reporting restless sleep most nights or every night was 59.9 years (standard deviation (SD): 8.5 years) and for those reporting <1 night a week with restless sleep it was 63.3 years (SD: 8.8). While 68.6% of those reporting restless sleep most or all of the time were women, 53.7% of those reporting rarely/or no restless sleep were women. While 30.7% of those reporting restless sleep most nights were non-Hispanic black, 21.8% of those reporting rarely or no restless sleep were non-Hispanic black. Distributions of education and income varied across frequency of restless sleep: attaining ≤ high school education occurred in 29.6% of those reporting most nights with restless sleep and 16.5% of those rarely/or never having restless sleep, while a reported income <$25,000 occurred in 24.8% of those reporting most nights with restless sleep and 14.2% of those rarely/or never having restless sleep. While 55.3% of those reporting restless sleep most nights the time had a body mass index ≥30 kg/m², 43.3% of those reporting rarely or no restless sleep had a body mass index ≥30 kg/m². Of those experiencing restless sleep most nights, 75.2% reported multi-joint symptoms and for those with rarely having restless sleep, 44.3% reported multi-joint symptoms. Nearly a third of those having restless sleep for most nights had depression, only 2.3% of those with rarely or never having restless sleep had depression.

Table 1.

Characteristics of participants by frequency of restless sleep at baseline (N=2,517).

Baseline characteristics	Rarely or never (<1 night) (n=995)	Sometimes (1–2 nights) (n=1,081)	Much of the time (3–4 nights) (n=288)	Most nights (5–7 nights) (n=153)
Mean (SD) age in years	63.3 (8.8)	62.7 (9.1)	61.1 (9.2)	59.9 (8.5)
Women (%)	53.7	58.6	64.2	68.6
Race/ethnicity (%)
Non-Hispanic White	74.8	79.6	79.5	67.3
Non-Hispanic Black	21.8	18.0	15.6	30.7
Other	3.4	2.4	4.9	2.0
Education (%)
High school or less	16.5	16.0	19.4	29.6
Some college	23.7	24.5	26.4	32.9
College graduate	21.4	21.6	17.4	17.1
Some graduate school or above	38.4	37.9	36.8	20.4
Income (%)
<$25,000	14.2	13.4	17.7	24.8
$25,000 – $50,000	27.9	28.5	25.7	30.1
>$50,000	57.9	58.1	56.6	45.1
Body mass index (%)
≤ 25 kg/m2	15.4	18.5	17.4	11.8
25–30 kg/m2	41.4	38.9	32.6	32.9
≥30 kg/m2	43.3	42.5	50.0	55.3
Kellgren-Lawrence grade (%)
2	56.1	56.1	61.1	63.4
3	32.6	32.9	30.2	30.7
4	11.4	11.0	8.7	5.9
Multi-joint symptoms (%)	44.3	52.6	63.2	75.2
History of knee injury (%)	37.0	41.2	38.9	43.8
History of knee surgery (%)	23.9	28.8	22.6	24.8
CES-D (>16) (%)	2.3	5.8	16.4	31.4
Comorbidity status
0	75.5	73.1	73.9	65.6
1	15.0	16.3	15.3	19.9
≥2	9.5	10.7	10.8	14.6
NSAIDs use	28.4	31.8	32.5	33.1
WOMAC scores, mean (SD)
Pain	3.2 (3.6)	3.9 (3.8)	4.8 (4.5)	5.9 (4.8)
Stiffness	1.9 (1.7)	2.2 (1.8)	2.5 (2.0)	2.8 (1.9)
Physical function	10.2 (11.5)	12.2 (12.3)	14.5 (13.8)	18.8 (15.6)
SF-12 PCS, mean (SD)	48.8 (8.9)	4.9 (8.7)	45.6 (10.2)	42.6 (12.0)
SF-12 MCS, mean (SD)	56.2 (6.3)	53.9 (7.6)	50.2 (9.5)	46.2 (12.0)
PACE, mean (SD)	156.0 (80.8)	157.6 (80.0)	158.7 (86.3)	147.9 (80.0)

Abbreviation: CES-D, Centers for Epidemiologic Studies Depression Scale; MCS, SF-12 Mental Component Summary scores; NSAIDs, non-steroidal anti-inflammatory drugs; PACE, Physical activity scale for the elderly; PCS, SF-12 Physical Component Summary scores; SD, standard deviation; WOMAC, The Western Ontario and McMaster Universities Arthritis Index.

Table 1 also shows the average symptoms and standard deviations (SD) across frequency of restless sleep, with adjusted estimates of the cross-sectional associations provided in Table 2. For those with rarely having restless sleep, the average pain, stiffness, and physical function scores were 3.2 (SD 3.6), 1.9 (SD 1.7) and 10.2 (SD 11.4), respectively. For those with restless sleep on most nights, the average pain, stiffness, and physical function scores were 5.9 (SD 4.8), 2.8 (SD 1.9) and 18.8 (SD 15.6), respectively. For both men and women, a linear trend (all p-values for linear trend < 0.01) between each symptom (pain, stiffness, and function) and frequency of restless sleep was observed and remained after adjusting for age, race/ethnicity, K-L grade, body mass index, and physical activity scores (Table 2). For example, in women with restless sleep occurring rarely as the reference group, the adjusted β coefficient for pain was 0.99 (95% confidence interval: 0.56 to 1.42) for 1–2 nights of restless sleep, 1.78 for 3–4 nights per week (95% confidence interval: 1.15–2.41), and 1.93 for most nights (95% confidence interval: 1.12 to 2.74). Similar patterns were observed for men.

Table 2.

Cross-sectional associations between restless sleep and knee symptoms.

	Rarely or never (<1 night)	Sometimes (1–2 nights)	Much of the time (3–4 nights)	Most nights (5–7 nights)	P-value for linear trend
Adjusted* β coefficient (95% Confidence Intervals).
All participants
Pain±	(Referent)	0.69 (0.39 to 1.00)	1.37 (0.90 to 1.84)	1.82 (1.20 to 2.45)	P<0.001
Stiffness±	(Referent)	0.28 (0.13 to 0.42)	0.48 (0.26 to 0.70)	0.55 (0.25 to 0.84)	P<0.001
Function±	(Referent)	2.08 (1.11 to 3.06)	3.43 (1.93 to 4.93)	5.49 (3.49 to 7.49)	P<0.001
Women
Pain*	(Referent)	0.99 (0.56 to 1.42)	1.78 (1.15 to 2.41)	1.93 (1.12 to 2.74)	P<0.001
Stiffness*	(Referent)	0.38 (0.18 to 0.58)	0.70 (0.40 to 0.99)	0.57 (0.19 to 0.94)	P<0.001
Function*	(Referent)	3.00 (1.62 to 4.37)	4.32 (2.30 to 6.33)	5.68 (3.09 to 8.27)	P<0.001
Men
Pain*	(Referent)	0.34 (−0.10 to 0.77)	0.78 (0.06 to 1.50)	1.85 (0.85 to 2.86)	P<0.001
Stiffness*	(Referent)	0.16 (−0.05 to 0.37)	0.17 (−0.18 to 0.42)	0.63 (0.15 to 1.12)	P=0.01
Function*	(Referent)	0.95 (−0.42 to 2.31)	2.27 (0.01 to 4.53)	5.89 (2.68 to 9.09)	P<0.001

^±Adjusted for age (linear term), sex, race/ethnicity, K-L grade, body mass index, CES-D, and physical activity scores.

^*Adjusted for age (linear term), race/ethnicity, K-L grade, body mass index, CES-D, and physical activity scores.

Descriptive analyses confirmed the time-varying nature of reported restless sleep. Approximately 28% reported rarely or never having restless sleep on each visit. Fifteen percent reported some nights having restless sleep on each visit in which a response was provided. Only 3 participants consistently reported restless sleep much of the time and 11 participants consistently reported restless sleep most nights on each visit.

Table 3 displays the results of the longitudinal analyses which treated restless sleep as the primary determinant of interest and each KOA symptom as the outcome. For women, no linear trend was observed for stiffness (p for linear trend = 0.16). Although in women the linear trend for pain was statistically significant (p for linear trend =0.01), the adjusted β coefficients for pain comparing each restless sleep category to rarely or no restless sleep were not suggestive of clinically important differences (most nights: −0.3 (95% CI: −0.61–0.0.01); 3–4 nights a week: −0.24 (95% CI: −0.49–0.02), 1–2 nights a week: −0.06 (95% CI: −0.2 to 0.08). Similar pattern was observed in physical function. For men, no linear trend was observed for pain (p for linear trend = 0.14) or stiffness (p for linear trend=0.97). Although the linear trend for function was statistically significant in men (p for linear trend = 0.04), the adjusted β coefficients for function comparing each restless sleep category to rarely or no restless sleep were not suggestive of clinically important differences (most nights: 0.84 (95% CI: −0.68–2.37); 3–4 nights a week: 0.8 (95% CI: −0.2–1.81), 1–2 nights a week: 0.45 (95% CI: −0.07 to 0.96).

Table 3.

Longitudinal associations between restless sleep and yearly changes in knee symptoms.

	Range of minimum clinically important difference (+:worsening; −: improvements)	Rarely or never (<1 night)	Sometimes (1–2 nights)	Much of the time (3–4 nights)	Most nights (5–7 nights)	P-value for linear trend
Adjusted* β coefficient (95% Confidence Intervals).
All participants
Pain±	1.2 to 4.6	(Referent)	0.004 (−0.1 to 0.1)	−0.05 (−0.24 to 0.14)	−0.16 (−0.39 to 0.08)	P=0.22
Stiffness±	0.5 to 1.5	(Referent)	0.002 (−0.05 to 0.05)	−0.01 (−0.10 to 0.08)	−0.08 (−0.19 to 0.04)	P=0.28
Function±	4.1 to 9.9	(Referent)	0.35 (−0.02 to 0.72)	0.25 (−0.37 to 0.88)	0.39 (−0.51 to 1.3)	P=0.16
Women
Pain*	1.2 to 4.6	(Referent)	−0.06 (−0.2 to 0.08)	−0.24 (−0.49 to 0.02)	−0.3 (−0.61 to 0.01)	P=0.01
Stiffness*	0.5 to 1.5	(Referent)	0.03 (−0.04 to 0.1)	−0.26 (−0.15 to 0.1)	−0.13 (−0.28 to 0.02)	P=0.16
Function*	4.1 to 9.9	(Referent)	0.26 (−0.25 to 0.77)	−0.13 (−0.93 to 0.68)	−0.01 (−1.14 to 1.12)	P=0.04
Men
Pain*	1.2 to 4.6	(Referent)	0.07 (−0.07 to 0.2)	0.22 (−0.05 to 0.49)	0.07 (−0.3 to 0.43)	P=0.14
Stiffness*	0.5 to 1.5	(Referent)	−0.03 (−0.11 to 0.04)	0.01 (−0.12 to 0.14)	0.03 (−0.15 to 0.21)	P=0.97
Function*	4.1 to 9.9	(Referent)	0.45 (−0.07 to 0.96)	0.8 (−0.2 to 1.81)	0.84 (−0.68 to 2.37)	P=0.04

^±Adjusted for age (linear term), sex, race/ethnicity, K-L grade, body mass index, CES-D, and physical activity scores.

^*Adjusted for age (linear term), race/ethnicity, K-L grade, body mass index, CES-D, and physical activity scores.

In Table 4, we show the results of the longitudinal analyses which restless sleep was treated as the outcome with rarely or no restless sleep treated as the reference group and each yearly change in KOA symptoms as the determinant. Yearly one-unit increases in pain scores was associated with decreased odds of having restless sleep most nights in women (aOR: 0.97; 95% CI: 0.93–1.00), much of the time (aOR: 0.97; 95% CI: 0.94–1.00), and some nights (aOR: 0.99; 95% CI: 0.97–1.01) relative to rarely/none of the time (<1 night). No association between pain and restless sleep was observed for men. Further, the data did not support an association between changes in symptoms such as stiffness and function and restless sleep in both women and men.

Table 4.

Longitudinal associations between yearly changes in knee symptoms and restless sleep.

	Rarely or never (<1 night)	Sometimes (1–2 nights)	Much of the time (3–4 nights)	Most nights (5–7 nights)
Adjusted* odds ratios coefficient (95% Confidence Intervals).
All participants
Pain±	(Referent group of the outcome variable)	1.00 (0.99 to 1.01)	0.99 (0.97 to 1.02)	0.98 (0.95 to 1.01)
Stiffness±	(Referent group of the outcome variable)	1.00 (0.97 to 1.02)	1.00 (0.95 to 1.04)	0.96 (0.91 to 1.02)
Function±	(Referent group of the outcome variable)	1.00 (1.00 to 1.01)	1.00 (1.00 to 1.01)	1.00 (0.99 to 1.02)
Women
Pain*	(Referent group of the outcome variable)	0.99 (0.97 to 1.01)	0.97 (0.94 to 1.00)	0.97 (0.93 to 1.00)
Stiffness*	(Referent group of the outcome variable)	1.01 (0.97 to 1.04)	1.00 (0.94 to 1.06)	0.95 (0.89 to 1.01)
Function*	(Referent group of the outcome variable)	1.00 (1.00 to 1.01)	1.00 (0.99 to 1.01)	1.00 (0.99 to 1.01)
Men
Pain*	(Referent group of the outcome variable)	1.00 (0.99 to 1.04)	1.04 (0.99 to 1.09)	1.01 (0.95 to 1.07)
Stiffness*	(Referent group of the outcome variable)	0.98 (0.94 to 1.02)	0.99 (0.92 to 1.07)	1.00 (0.90 to 1.12)
Function*	(Referent group of the outcome variable)	1.01 (1.00 to 1.01)	1.01 (1.00 to 1.03)	1.01 (0.99 to 1.03)

^±Adjusted for age (linear term), sex, race/ethnicity, K-L grade, body mass index, CES-D, and physical activity scores in a multinomial model with restless sleep included as the outcome variable and separate models for each KOA symptom as the primary determinant of interest.

^*Adjusted for age (linear term), race/ethnicity, K-L grade, body mass index, CES-D, and physical activity scores.

Discussion

Using cross-sectional and longitudinal data from the OAI, this study adds to the current literature on frequency of restless sleep and KOA-related symptoms. The OAI is a multi-site, longitudinal study that offers a large sample size of participants with radiographically confirmed knee OA and validated patient reported symptoms of KOA. Analysis of cross-sectional data supported a strong linear association between frequency of restless sleep and KOA pain, stiffness, and function symptoms, with similar trends observed for men and women. These associations did not uniformly hold-up in the more robust longitudinal design.

Sleep and pain are inter-related conditions. Of those with chronic pain, more than half report sleep complaints [35]. Among people reporting chronic insomnia, half report pain [36]. What is less clear is the direction of this association and the mechanisms through which the associations occur. In addition to a cross-sectional relationship, our findings suggest that KOA symptoms may be associated with frequency of restless sleep, rather than restless sleep impacting severity of symptoms including pain, stiffness, and function. Experimental evidence in healthy volunteers demonstrated that sleep disruption was associated with self-reported morning aching or stiffness [37]. Experimental evidence in patients with rheumatoid arthritis linked partial sleep disruption to symptoms in the morning after as documented by both clinicians (joint counts) and patients (self-reported pain) [38]. None of these studies specifically evaluated patients with KOA, nor were the studies of sufficient duration to be applicable to KOA sufferers. Using accelerometer data collected in the OAI, poor sleep quality was associated with less physical activity in adults with or at risk for KOA [17]. It may be that the sleep-pain association was masked in our study because those with poor sleep quality avoided potentially painful activities brought about by physical activity. Future work should consider this possibility.

Our findings are not consistent with evidence supporting the notion that the association between sleep and pain is bidirectional. Prolonged sleep restriction has been shown to stimulate pro-inflammatory responses (e.g., elevation of circulating IL-6) [39]. Such findings support the notion that sleep quality may impact KOA symptoms; we were unable to explore the role of inflammation as we did not have data on inflammatory biomarkers. Further, a meta-analysis based on a limited number of studies suggested that sleep interventions may hold promise to alleviate symptoms in patients with chronic pain [19]. Despite that our findings support the sleep quality → pain and other symptoms pathway, our analyses in participants with radiographically confirmed KOA did not suggest that pain and other KOA symptoms may impact frequency of restless sleep. Future research further characterize such bidirectional associations may be warranted.

We believe that our longitudinal finding suggesting no clinically meaningful association between restless sleep and symptoms of KOA is important. Taken with the results of the cross-sectional analysis, our findings suggest that KOA symptoms may impact sleep quality. Because treatment goals for KOA focus on improving quality of life, broadening the scope of recommendations to include suggestions for improving sleep for those with KOA symptoms is logical but warrants further investigation.

The strengths and limitations of this study should be kept in mind when reviewing these findings. The OAI offers a large sample of community-dwelling individuals with radiographically confirmed KOA. The OAI offers a vast array of information regarding patient-reported KOA symptoms and potentially confounding factors. We were able to conduct two separate analyses using these data: 1) cross-sectional at baseline; 2) longitudinal analyses allowing for restless sleep to vary with time. Together, these analyses allow for comparison with previous research. A limitation of our research is that the only assessment of sleep available in the OAI data was the single item from the CES-D scale on restless sleep. To what extent this single item is associated with sleep quality remains unclear and the measurement error of using a single sleep measure could be a concern. Despite that the sleep measure from CES-D may not be an adequate substitute for a multi-item measure of overall sleep quality, it has been used to assess sleep in prior studies and shown to yield good variability both between and within participants [21, 22, 40]. Future work using multi-item measures including other aspects of sleep quality, validated sleep scales, or objective measures of sleep disturbance in relation to self-reported KOA symptoms are warranted. In addition, we believe that the seemingly contradictory findings from cross-sectional and longitudinal analyses are most likely due to residual confounding, such as fatigue [41]. Despite the comprehensive information provided in OAI, the assessment regarding fatigue and/or biomarkers that are potentially associated with poor sleep are lacking. On the other hand, fatigue could also be an underlying factor that mediates the association of knee symptoms and sleep. Future research fully evaluating the extent to which knee symptoms and sleep are mediated by factors such as fatigue are needed. Lastly, the generalizability of our study findings may be limited. The OAI sample was predominantly non-Hispanic white, more educated, and higher income. In addition, OAI is a community-based sample which participants may show little symptom progression over time compared to healthcare seekers.

In conclusion, the linear trend between frequency of restless sleep and self-reports of KOA pain, function, and stiffness reports among men and women observed in cross-sectional analyses were not found in the more rigorous longitudinal analysis. Although in women, a statistically significant linear trend was observed for the association between frequency of restless sleep and pain and physical function. In men a statistically significant linear trend was observed for the association between frequency of restless sleep and function. Neither appeared to reach the a priori selected ranges for minimally clinically relevant differences. Taken together, our findings support the contention that KOA symptoms affect frequency of restless sleep, rather than the reverse. There is a known relationship between poor sleep and 1) mortality [42, 43], 2) cardiovascular events [43], and 3) other adverse health outcomes [44, 45]. Given the prevalence of frequent restless sleep and the impact of KOA symptoms on sleep observed in this study, treatment goals focusing on improving patients’ quality of life should also consider sleep hygiene interventions as symptoms such as pain, stiffness, and physical functioning are addressed.

Key Points:

• The prevalence of frequent restless sleep among persons with knee OA is not uncommon.

• There were linear trends between frequency of restless sleep and self-reported symptoms of the knee in cross-sectional analyses.

• In the more robust longitudinal analysis, despite the statistically significant linear trends observed between frequency of restless sleep and symptoms (women: pain and physical function; men: function), none appeared to reach the a priori selected ranges for minimally clinically relevant differences.