Abstract
Objective The aim of this systematic review and metaanalysis is to summarize evidence regarding the relationship between psoriatic arthritis (PsA) and sleep problems.
Methods We identified 36 eligible studies—26 cross-sectional, 7 cohort, and 3 interventional studies—in PubMed and Embase.
Results The prevalence of self-reported sleep problems in patients with PsA ranged from 30% to 85%. A metaanalysis of 6 studies that used the Pittsburgh Sleep Quality Index revealed a prevalence of poor sleep quality for patients with PsA of 72.9% (95% CI 63-81.8; I2 = 78%), which was statistically higher than in healthy controls (26.9%, 95% CI 11.7-45.4; I2 = 81%) but not significantly different than in patients with psoriasis (59.8%, 95% CI 46.9-72.1; I2 = 51%). Sleep disturbance was ranked in the top 4 health-related quality of life domains affected by PsA. One study suggested a bidirectional relationship between PsA and obstructive sleep apnea. Predictors of sleep problems included anxiety, pain, erythrocyte sedimentation rate, depression, fatigue, physical function, and tender or swollen joint count. Tumor necrosis factor inhibitors, guselkumab, and filgotinib (a Janus kinase inhibitor) were associated with improved sleep outcomes.
Conclusion Poor sleep quality is prevalent in patients with PsA. Objective sleep measures (ie, actigraphy and polysomnography) have not been used in PsA studies, and evidence on the validity of patient-reported sleep measures in PsA is lacking. Future studies should validate self-reported sleep measures in PsA, explore how sleep quality relates to PsA disease activity and symptoms using both objective and subjective sleep measures, assess the efficacy of strategies to manage sleep problems, and assess the effects of such management on symptoms and disease signs in patients with PsA.
Psoriatic arthritis (PsA) is a complex inflammatory disease that manifests as peripheral arthritis, dactylitis, enthesitis, and/or spondylitis.1 PsA is diagnosed in approximately 30% of patients with skin psoriasis (PsO)2 and is associated with other systemic inflammatory diseases, including inflammatory bowel disease, cardiovascular disease, and type 2 diabetes.3
PsA has also been associated with sleep problems, including short sleep duration, poor sleep quality, and sleep disorders. Sleep is vital for proper physiologic and psychologic functioning.4-6 Sleep problems are associated with chronic physical and mental health problems, including cardiovascular disease, type 2 diabetes, obesity, and depression. Multiple studies have also shown diminished quality of life in patients who have sleep problems.7,8 Sleep problems, therefore, may present a potential additional source of risk—and of intervention—for patients with PsA who are already at risk for such comorbidities.
A growing body of literature has investigated the relationship between psoriatic disease (PsD) and sleep. The aim of this systematic review is to synthesize available evidence regarding the (1) prevalence and incidence of sleep problems in patients with PsA, (2) prevalence and incidence of PsA in patients with sleep problems, (3) factors associated with sleep problems in patients with PsA, (4) use of validated vs unvalidated sleep outcome measures in patients with PsA, (5) beliefs about sleep in patients with PsA, and (6) effect of PsA therapies on sleep problems in patients with PsA.
METHODS
This review was conducted in accordance with the Preferred Reporting Items for Systematic reviews and Meta-Analyses guideline.9 All definitions and abbreviations used in this study are found in Supplementary Table S1 (available with the online version of this article).
Inclusion criteria. We included studies that were full-text original articles in English featuring older adolescents and adults (≥ 16 years) in which any of the following conditions were met: (1) sleep problems were evaluated as an outcome in patients with PsA, (2) the incidence or prevalence of PsA was evaluated in patients with sleep disorders; (3) the quality (ie, measurement properties) of sleep measures was evaluated in PsA, and (4) beliefs about sleep were elicited by patients with PsA. Subanalysis of sleep problems for the PsA population was required for studies including other populations (eg, PsO and ankylosing spondylitis).
Exclusion criteria. We excluded any non-English–language studies, review articles, research letters, opinion letters, notes, abstracts, and animal studies.
Literature search. PubMed and Embase databases were searched on January 5, 2023. The PICOS (population, intervention, comparison, outcome, and study design) criteria, and Emtree and MeSH terms were used to create a thorough search strategy (Supplementary Table S2, available with the online version of this article). Search results were uploaded to Covidence and reviewed first by title and abstract followed by full-text review. Studies were chosen for inclusion by 3 independent reviewers (CG, MW, and LMPC).
Data extraction. Data extraction was performed by each independent reviewer with subsequent discussion about any disagreements. Data were organized per study aim: (1) prevalence and incidence of sleep problems in patients with PsA, (2) prevalence and incidence of PsA in patients with sleep problems, (3) factors associated with sleep problems in patients with PsA, (4) use of validated vs unvalidated sleep outcome measures in patients with PsA, (5) beliefs about sleep in patients with PsA, and (6) effect of PsA therapies on sleep problems in patients with PsA.
For the purpose of this work, we agreed that an outcome measure would be considered as validated if any of the following measurement properties was assessed in a validation study for any population: content validity, construct validity, structural validity, reliability, and responsiveness. If none of these measurement properties was assessed in any population, the instrument was deemed as unvalidated.
Quality evaluation. Two independent reviewers (LMPC and MW) evaluated the quality of the data using the Standard Quality Assessment Criteria for Evaluating Primary Research Papers from a Variety of Fields developed by Kmet et al.10 This quality assessment tool was selected given the heterogeneity of study designs included in the review. Studies were classified as being of high quality (> 8 out of 10), good quality (6-8 out of 10), moderate quality (4-6 out of 10), or poor quality (< 4).11
Statistical analysis. Proportional metaanalyses were implemented to evaluate the pooled prevalence of poor sleep quality as determined by the Pittsburgh Sleep Quality Index (PSQI).12 Heterogeneity between studies was evaluated with I2 statistics and considered present if I2 > 50% (P < 0.1).13 The random effects model was applied given that the I2 statistics indicated high heterogeneity. Additional metaanalyses with single means and SDs were incorporated to examine pooled PSQI variables (ie, subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, use of sleeping medication, and daytime dysfunction). An individual metaanalysis was performed on patients with PsA, patients with PsO, and healthy controls (HCs) if at least 2 studies were available.14 The assessment of publication bias was not considered relevant during the proportional metaanalyses,15 and not relevant during the metaanalyses with single means because of the low number of included studies. The statistical analysis was conducted using the statistical software R (R Foundation for Statistical Computing) with the additional meta package.
RESULTS
In all, 810 references were retrieved; duplicates were removed, resulting in 719 unique references for screening (Figure 1). After screening the title and/or abstract, 152 articles remained for full-text assessment. There were 116 full-text studies excluded. A total of 36 articles were included (Table 1).16-51
PRISMA flow diagram. PRISMA: Preferred Reporting Items for Systematic reviews and Meta-Analyses.
Characteristics of included studies.
Study design and quality evaluation. We identified (1) 33 observational studies, including 26 cross-sectional studies16-41 and 7 cohort studies,42-48 and (2) 3 interventional studies, including 2 randomized controlled trials (RCTs)49,50 and 1 non-RCT.51 Supplementary Table S3 (available with the online version of this article) includes the quality evaluation for the studies. Most studies were of high quality (ie, low risk of bias). Common limitations of these studies included absence of sample size calculations and limited methods to adjust for confounders. Two qualitative studies26,45 had moderate risk of bias because of limitations in sampling strategy, verification of study results, and data collection methods.
Prevalence and incidence of sleep problems in patients with PsA. Sleep problems were defined in different ways across studies, using a variety of measures. Based on these multiple definitions, the prevalence of sleep problems in patients with PsA ranged from 30% to 85% (Table 1 and Supplementary Table S4, available with the online version of this article). A metaanalysis of the prevalence of poor sleep quality across 6 studies20,21,28,36-38 that used the PSQI12 revealed a prevalence of poor sleep quality for patients with PsA of 72.9% (95% CI 63-81.8; I2 = 78%), which was statistically higher than in HCs (26.9%, 95% CI 11.7-45.4; I2 = 81%)28,37,38 but not significantly different than in patients with PsO (59.8%, 95% CI 46.9-72.1; I2 = 51%)21,28,37,38 (Figure 2 and Supplementary Table S5, available with the online version of this article).
Metaanalysis of the prevalence of poor sleep quality as measured by the PSQI in patients with psoriatic arthritis, patients with psoriasis, and healthy controls. PSQI: Pittsburgh Sleep Quality Index.
Regarding discrete sleep disorders, as defined by the International Classification of Sleep Disorders, 1 study reported that the prevalence of restless legs syndrome (RLS) in patients with PsA was 64%, which was significantly higher compared to patients with PsO (20%) and HCs (14%).33 Insomnia was more frequent in patients with axial spondyloarthritis (axSpA) than in patients with PsA (PsA insomnia score: median 26.0, IQR 18.0-30.7, vs axSpA insomnia score: median 22.0, IQR 15.2-28.7; P = 0.014); prevalence estimates, however, were not reported18 (Table 1 and Supplementary Table S4, available with the online version of this article).
Using data from the Danish National Patient Register, Egeberg et al43 reported an incidence rate ratio (IRR) of 1.75 (95% CI 1.35-2.26) for sleep apnea (International Classification of Diseases, 10th revision, diagnosis code G47.3) in patients with PsA, after adjusting for age, sex, smoking, alcohol, comorbidities, and socioeconomic status (Table 1 and Supplementary Table S6, available with the online version of this article).
Prevalence and incidence of PsA in patients with sleep apnea. Egeberg et al43 identified a high incidence of PsA in patients with sleep apnea (IRR of PsA in sleep apnea treated with continuous positive airway pressure [CPAP]: 1.94, 95% CI 1.34-2.79, and without CPAP: 5.59, 95% CI 3.74-8.37), showing a significant bidirectional association in adjusted models. Cohen et al42 further reported that patients with obstructive sleep apnea (OSA) had an increased risk of developing PsA (Table 1 and Supplementary Table S6, available with the online version of this article). No studies determining the prevalence and incidence of PsA in other sleep disorders (eg, insomnia and RLS) were found.
Factors associated with sleep problems in patients with PsA. Overall, we found moderate to strong (0.3 ≤ r < 0.7) correlations between sleep problems and pain, fatigue, physical function, emotional distress, tender joint count (TJC), enthesitis, erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), patient age, and number of tender joints (Supplementary Table S7, available with the online version of this article).
In studies in which the sample consisted of patients with PsO and PsA, PsA was identified as an independent predictor of sleep problems.17,34,51 Other factors independently associated with sleep problems in this population were itch, physical pain or soreness, female sex, obesity, OSA, moderate and severe PsO, smoking, poor dermatologic quality of life, and work productivity impairment (Supplementary Table S8, available with the online version of this article).
In studies in which the sample consisted of patients with PsA (with or without PsO),18,21,23,28,31,37,38 pain was the most frequently identified factor independently associated with sleep problems, followed by actively inflamed or tender joints. Other factors independently associated with sleep problems included depression, anxiety, emotional recovery, physical function, itch, ESR, CRP, age, and duration of PsO (Supplementary Table S8, available with the online version of this article).
Sleep problems were identified as a predictor of fatigue52 (ie, a sustained, overwhelming sense of exhaustion associated with decreased capacity for physical and mental work that is different from sleepiness, that is, a propensity to sleep53). Sleep problems were also identified as a predictor of poor health-related quality of life, as measured by the 15-Dimensional questionnaire, and active PsA, as measured by the Disease Activity Index for Psoriatic Arthritis29 (Supplementary Table S8, available with the online version of this article).
Use of validated sleep PROMs in patients with PsA. In total, 24 studies (67%) used a patient-reported outcome measure (PROM) to assess sleep (Table 1). Six of these studies used a sleep question that had not been previously validated in any population. Six studies used a sleep question embedded in a validated PsD PROM (ie, Psoriatic Arthritis Impact of Disease [PsAID], Routine Assessment of Patient Index Data 3, and Multidimensional Health Assessment Questionnaire). In total, 12 studies used a PROM that was previously validated to measure sleep in other populations but not in PsA (ie, Ottawa Sleep Scale, PSQI, Patient-Reported Outcomes Measurement Information System [PROMIS] Sleep Disturbance, Epworth Sleepiness Scale, Insomnia Severity Index, and Medical Outcomes Sleep Scale; Supplementary Table S9, available with the online version of this article).
We identified 4 studies evaluating 1 or more measurement properties of sleep measures in patients with PsA.22,29,41,44 The evaluated sleep measures included the Jenkins Sleep Scale (JSS), the PsAID item sleep disturbance, the Assessment of SpondyloArthritis international Society Health Index item I sleep badly at night, and the sleep visual analog scale (VAS: 0-100 mm; 0 mm = no/minimal symptoms, 100 mm = worse symptoms; Table 1).
Sleep problems as measured by the PSQI. Eight studies assessed sleep problems in patients with PsA with the PSQI.20,21,28,33,35-38 PSQI scores ranged from 5.0 to 10.0. Patients with PsA consistently presented with worse sleep quality than HCs and patients with PsO. Patients with PsA who had neuropathic pain or fibromyalgia had worse sleep quality than patients without neuropathic pain or fibromyalgia.35,36
A metaanalysis of the mean scores for the sleep dimensions assessed by the PSQI (ie, PSQI variables) across 3 studies showed that mean sleep quality, latency, disturbances, and day dysfunction were significantly higher in patients with PsA compared to HCs (Supplementary Table S10 and Supplementary Figures S1-S4, available with the online version of this article).21,28,37 Evaluating the overlapping CI, differences between patients with PsA and PsO were considered not significant. Use of sleep medication was comparable across groups. PSQI variables are each scored on a 0 to 3 scale, where a higher score reflects worse sleep. Sleep efficiency was the sleep dimension most affected (mean 2, 95% CI 1.1-2.9), followed by sleep latency (mean 1.6, 95% CI 1.3-1.9), sleep disturbance (mean 1.6, 95% CI 1.4-1.8), and day dysfunction (mean 1.5, 95% CI 1.1-2; Supplementary Table S10).
Sleep problems as measured by the PsAID questionnaire. Four studies evaluated sleep with the sleep disturbance item from the PsAID questionnaire.19,31,32,49 The score ranged from 0 to 7, with a mean ranging from 1.9 to 3.4 (SD 2.4-3.3). In Palominos et al,31 a score of 4 or higher was considered as having sleep disturbance mirroring the PsAID cut-off value used to identify patients at a patient acceptable symptom state (PASS). Other studies did not clarify what cut-off value was used to define sleep disturbance. Puyraimond-Zemmour et al32 reported that a value of 2 for the “sleep disturbance” item corresponded to the PASS level. This threshold had a sensitivity of 0.77 and specificity of 0.50 (Table 1).
Studies evaluating beliefs about sleep problems in patients with PsA. Six studies explored patients’ beliefs about sleep problems in patients with PsA (Table 1).26,27,30,39,40,45 Hu et al26 reported that 60% of participants believed that sleep was affected by PsA and ranked sleep disturbances in the top 4 domains affected by PsA. Additionally, they reported that 59% of participants were willing to pay for a cure in this domain (median $10,000, IQR $5000-$50,000).26 In Nowell et al,45 patients with PsA ranked the PROMIS Sleep Disturbance domain as the fourth-most important PROM to track disease management. In 1 study, 78 (67%) participants believed that biologics had a “very beneficial” or “beneficial” effect on sleep quality27; in another study, 29 (63%) patients believed that the use of cannabis helped them sleep better.40 More that 80% of participants further identified sleep disturbance as an important effect in patients with PsA, along with physical disability, effects on daily activities, and feelings of frustration.30 In a study involving 332 participants, sleep quality was identified as the third-most important disease effect that required improvement, ranking higher than the ability to perform physical activities, participate in work, take part in social or leisure activities, and emotional well-being.39
Effect of PsA therapies on sleep problems in patients with PsA. Three studies, including 1 non-RCT,51 1 cohort study,47 and 1 cross-sectional study,28 showed that tumor necrosis factor (TNF) inhibitors improved sleep problems in patients with PsA as measured by the PSQI and the Medical Outcomes Study Sleep Scale. One cohort study showed improvement in sleep quality (VAS) with ustekinumab,48 but changes were not statistically significant, whereas an RCT reported improved sleep with guselkumab.50 Another RCT showed significant improvement in the PsAID sleep disturbance item following treatment with filgotinib, a Janus kinase (JAK) inhibitor.49 In a cohort study including initiators of TNF and interleukin 17 inhibitors, the proportion of patients who had poor sleep quality at baseline dropped from 71.9% to 47.4% (P = 0.01) by month 4. However, mixed effect models found that initiating either of those therapies did not influence changes in PSQI scores38 (Table 2).
Effect of therapies on sleep problems.
DISCUSSION
In this systematic review, we found that 30% to 85% of patients with PsA reported sleep problems. The difference in the rates of sleep problems across studies may be partly a result of the different sleep problems evaluated and the instruments used to measure them. The sleep problems evaluated included sleep disturbance, sleep interference, sleep difficulty, sleep impairment, sleep disorders, and poor sleep quality. Even among studies measuring the same sleep problem, different measurement instruments were used. For example, sleep disturbance was measured with an unvalidated numeric rating scale question, with the sleep disturbance question from the PsAID and with the sleep disturbance subscore from the PSQI. The lack of a common terminology and common measurement instrument for a given sleep problem across studies hinders comparison of results.
Despite the variability in sleep problems evaluated and sleep measures used among studies, we identified 6 studies reporting on the prevalence of poor sleep quality as measured by the PSQI.20,21,28,35,37,38 Our present metaanalysis based on these studies concluded that 73% of patients with PsA had poor sleep quality. Pooled data from 4 of these studies21,28,37,38 further showed that the prevalence of poor sleep quality in patients with PsA was significantly higher compared to HCs (27%). Compared to patients with PsO, pooled data from 3 of these studies28,37,38 revealed that differences were not significant. Additionally, our metaanalysis of mean scores of PSQI variables found that patients with PsA had worse subjective sleep quality, latency, disturbances, and day dysfunction than HCs, but not compared to patients with PsO. Results of these metaanalyses, however, are limited by the large heterogeneity identified and the small number of studies included.54 Differences in patient demographics and disease severity, the presence of comorbidities including sleep disorders, and medication status may account for the variability of results among studies.
In this review, only self-reported sleep measures were used; objective sleep measures (ie, actigraphy or polysomnography) and sleep diaries were not used in PsA. Henry et al55 described similar findings in their review for patients with PsO. Although they identified a few studies that used polysomnography in patients with PsO, actigraphy and sleep diaries were not applied. Both subjective and objective sleep measures are required to identify and assess sleep problems,56 especially for targeting interventions. Future studies in PsD should include both types of sleep assessments.
The JSS was the only sleep PROM whose measurement properties were formally assessed in a validation study. However, we did not find studies reporting results for the JSS in patients with PsA.41 In this validation study, investigators explored the internal consistency and construct validity of the JSS, but its content validity in PsA has not been explored. The PSQI is the most widely used sleep PROM in clinical and nonclinical populations. It was developed for patients with depression and later validated in several other disorders.57 Yet, the PSQI has not been validated in patients with PsA. Further, even in other populations, its dimensionality (ie, factor structure) warrants further investigation.58 Given that patients with psoriatic disease were not involved in the development of these instruments and that such instruments do not quantify symptoms that may cause sleep problems in these patients, such as pruritus, skin pain, joint pain, and stiffness, evaluation of the content validity and other measurement properties of sleep PROMs in PsA is warranted. To address the potential lack of comprehensiveness of these measures in PsD, a set of PsD–specific sleep measures—the PsO Sleepy-Q and the PsA Sleepy-Q—are currently being validated.59
This review identified several predictors of sleep problems in patients with PsA, including anxiety, pain, ESR, emotional recovery, depression, fatigue, physical function, and TJC or swollen joint count (SJC). Yet, the directionality of the association between sleep and these predictors has not been established. Longitudinal studies that include both PROMs and actigraphy are needed to explore whether the symptoms and/or signs of PsA (eg, joint pain or TJC/SJC) hold a bidirectional relationship with sleep problems as suggested in qualitative studies.59,60
Finally, we found that TNF inhibitors, guselkumab, and filgotinib were associated with improved sleep outcomes. However, these assessments presented certain limitations. First, only 249,50 of 728,38,47-51 studies consisted of an RCT. Second, sleep was not the primary outcome in any of the 2 cited RCTs.49,50 Therefore, power calculations were not based on sleep outcomes. Third, the sleep PROMs that were used had not been validated in PsA, and objective sleep measures were not used, as previously noted. Properly powered clinical trials exploring the efficacy of (1) PsD interventions and (2) sleep-directed interventions (eg, cognitive behavioral therapy for insomnia) to treat sleep outcomes in PsA are required.
This systematic review has limitations. Our search was restricted to 2 databases, PubMed and Embase, and inclusion criteria were restricted to data from individuals 16 years of age and older, studies evaluating sleep in patients with PsA (or at least subanalysis for the PsA population was required), and studies in English. Therefore, it is possible that studies outside of these parameters with information relevant to our study aims were not included.
In conclusion, sleep problems are common in patients with PsA, and patients consider sleep problems as having an important effect on PsA. The prevalence of poor sleep quality was significantly higher in patients with PsA compared to HCs, but not compared to those with PsO. Our results rely on a small number of heterogenous studies. Sleep apnea holds a bidirectional relationship with PsA in adjusted models, but data regarding the association between PsA and other sleep disorders is lacking. The sleep disturbance items from the PSQI and the PsAID were the most widely used sleep PROMs across studies, though evidence supporting their validity is needed. Longitudinal studies combining daily measurement of PsA signs and symptoms, subjective sleep measures (ie, validated sleep PROMs and sleep diaries), and objective sleep measures (eg, actigraphy) are needed to further understand how sleep quality relates to PsA disease activity and symptoms. Finally, although a few studies support the beneficial effect of biologics and JAK inhibitors for PsA on sleep problems, further RCTs evaluating the efficacy of diverse therapies to manage and improve sleep in PsA are needed.
Footnotes
EBK received support from the Leducq Trans-Atlantic Network of Excellence on Circadian Effects in Stroke, the National Institutes of Health (grants R01HD107064, R01NS114526-02S1, R21DA052861, U54-AG062322, and U01NS114001), and the US Department of Defense (grant W81XWH201076). LMPC received support from the National Psoriasis Foundation (grant 815789).
C. Grant and M. Woodbury contributed equally to this work as first authors.
EBK has consulted for the American Association of Sleep Medicine Foundation, the National Sleep Foundation, Circadian Therapeutics, and Yale University Press. JFM has served as a consultant and/or investigator for Dermavant, Eli Lilly, UCB, Sun Pfizer, and Leo Pharma. AO has consulted for AbbVie, Amgen, BMS, Celgene, CorEvitas, Gilead, GSK, HappifyHealth, Janssen, Novartis, Pfizer, and UCB; and received grants from AbbVie, Amgen, Novartis, and Pfizer. The remaining authors declare no conflicts of interest relevant to this article.
- Accepted for publication April 6, 2023.
- Copyright © 2023 by the Journal of Rheumatology
