To characterize sleep complaints in women with systemic lupus erythematosus (SLE) and to identify correlates of sleep quality.
To characterize sleep complaints in women with systemic lupus erythematosus (SLE) and to identify correlates of sleep quality.
Sleep quality in 100 women with SLE was assessed using the Pittsburgh Sleep Quality Index (PSQI). Participants completed standardized questionnaires assessing depressed mood, leisure time physical activity, functional disability, and pain severity. A clinical examination determined disease activity, cumulative damage, and whether patients fulfilled the American College of Rheumatology criteria for fibromyalgia. A series of hierarchical multiple regressions were computed.
The mean ± SD global PSQI score was 6.98 ± 4.03, with moderate to severe sleep impairment reported by 56% of the sample. The first model testing the importance of demographic factors was not statistically significant. In the disease-related model, the use of prednisone and functional disability both contributed to poor sleep quality (P < 0.001). The addition of level of exercise participation to the demographic set significantly added to the model (P = 0.001). Depression significantly added to the demographic set, explaining 29% of the variance (P < 0.0001). When these variables, along with disease related variables, were simultaneously regressed on the PSQI Global Score, only depressed mood appeared as a significant independent determinant of global sleep quality (P < 0.001). However, the point estimates for the Beta coefficients were consistent with effects for lack of exercise and prednisone use.
A significant proportion of women with SLE suffer from poor sleep quality. The findings suggest that depressed mood, prednisone use, and lack of exercise contribute to decreased overall sleep quality.
Difficulty initiating sleep, maintaining sleep, and/or early morning awakenings are all components of sleep disturbance. Sleep disturbances are relatively common in women, in older persons, and in individuals with fewer socioeconomic resources (1, 2). Sleep disturbance is often seen in rheumatic diseases, and may contribute to the fatigue that is part of many rheumatic conditions, including systemic lupus erythematosus (SLE). The prevalence of sleep impairment and the factors contributing to sleep quality in patients with SLE remains poorly understood. To our knowledge only 3 studies have examined sleep quality in SLE.
McKinley et al (3) found that SLE patients reported more sleep problems compared with healthy controls. Valencia-Flores et al (4) evaluated polysomnographic data in 14 SLE patients and found that sleep was characterized by respiratory (50%) and movement (50%) disorder. Disease activity was associated with more sleep fragmentation and with less sleep efficiency. More recently, a study comparing SLE patients with healthy controls found SLE patients reported more pain when trying to fall asleep and during the night (5). While the frequency of nocturnal awakenings was similar in both groups, SLE patients, when awakened, were awake for longer periods (>30 minutes) during the night.
The few studies conducted to date suggest that sleep disturbance is prevalent among patients with SLE. However, limitations of these studies include small sample sizes (ranging from 14–48 patients) (3–5) and the use of different sleep assessments (some unstandardized) (3). Although sleep disturbance is believed to be multidetermined (1), the specific factors involved and their relative importance in SLE remains virtually unknown. Although disease activity and pain intensity have been associated with sleep disturbance in SLE, little is known about the contribution of other factors (i.e., depression, exercise) associated with sleep quality in other chronic medical conditions.
The goals of the present study were to describe the sleep quality using the Pittsburgh Sleep Quality Index (PSQI) (6) in women with SLE, and to delineate factors associated with sleep quality in patients with SLE. We were guided by a biopsychosocial model of sleep that included variables such as disease activity and pain severity that have been identified in epidemiologic sleep studies, and that may be relevant in SLE.
The sample was comprised of patients who fulfilled at least 4 of the 1997 American College of Rheumatology (formally American Rheumatism Association) revised criteria for SLE (7, 8), were at least 18 years of age, functionally fluent in English or French, and had no major cognitive deficits that would preclude questionnaire completion. Because the vast majority (>90%) of persons with SLE are women, for clarity and feasibility, we studied women only.
Physicians invited consecutive patients with SLE, during their scheduled appointment at The McGill University Health Centre Lupus Clinic, to participate in the study. The research assistant obtained informed consent and reviewed the questionnaire protocol with the patient during the clinic visit. Patients were provided with a preaddressed stamped envelope to return the self-report questionnaires through the mail. Patients underwent a standard medical examination at the time of study entry. The study was approved by the McGill University Health Centre Institutional Review Board.
The PSQI (6) is a self-report measure that assesses sleep quality and disturbances over a 1-month interval. It includes 19 items, generating 7 component scores: subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, use of sleeping medication, and daytime dysfunction. A global score is obtained by summing the 7 component scores and has a possible range of 0–21 points. The scale has good psychometric properties with a global score ≥6 yielding a diagnostic sensitivity of 89.6% and specificity of 86.5% in differentiating good and poor sleepers (6).
The Center for Epidemiological Studies-Depression Scale (CES-D) is a psychometrically sound 20-item measure designed to assess depressive symptoms in nonpsychiatric populations (9–11). It has been widely used in chronic medical diseases (12). Previous work using the CES-D in persons with rheumatoid arthritis has shown that 4 items (items 7, 8, 11, and 20) may be more associated with arthritis than depression (13). Since inclusion of these items might thus confound the relationship between depression and sleep quality in our patients, these items were not included in the total score. The remaining 16 items were summed and multiplied by a constant of 1.25 to retain the original 0–60 range (CESD-AR) (13).
The Aerobics Center Longitudinal Study Physical Activity Questionnaire (ACLS-PAQ) is a brief validated instrument assessing participation in leisure and household activities in the last 3 months (14). A total physical activity score, and a total physical activity score excluding household activities, stair climbing, and lawn work can be expressed. These scores are estimates of weekly energy expenditure expressed as MET-hours per week. The intensity of each reported activity is converted to a MET value by using various compendia of physical activity (15–17). The ACLS-PAQ has been well validated and correlates well with objective measures of physical fitness (17–19).
The Functional Disability Index of the Health Assessment Questionnaire (HAQ) (20) was used to measure difficulty in performing activities of daily living. Scores on the disability index range from 0 to 3, with higher scores indicating more disability. The HAQ is a reliable and valid measure of functional disability in SLE (21, 22). Pain was measured by the HAQ 15-cm visual analog scale, with 0 denoting no pain and 100 indicating very severe pain. Pain scores were then rescaled from 0–3, with 0 representing no pain and 3 severe pain.
We used the Systemic Lupus Activity Measure-Revised (SLAM-R) (23), a reliable and validated instrument, to measure SLE disease activity over the past month. The SLAM-R is based on physician examination and laboratory assessment, which includes a complete blood cell count, erythrocyte sedimentation rate, creatinine clearance, and urinalysis. Scores may range from 0 (no disease activity) to 84 (maximum disease activity). Based on a study published by our group (24), a score >8 indicates moderate to severe clinical activity.
Cumulative damage was measured using the Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index (SLICC/ACR DI) (25, 26). The SLICC/ACR DI is a physician-rated index that assesses cumulative organ damage due to either the disease, complications of therapy, or intercurrent illness such as diabetes or cancer. Total scores on this index range from 0 (no damage) to 46 (maximum damage).
Sociodemographic and additional clinical information including disease duration, age, education, and marital and employment status were collected by self report. Participants also reported on prescribed medications.
Descriptive statistics including means, medians, and standard deviations were calculated for all the variables. Univariate outliers were identified for the variable pertaining to weekly MET-hours related to leisure physical activity. Following the recommendations of Tabachnick and Fidell (27) for handling continuous variables with univariate outliers, the data were grouped on a scale of 1 to 3 (1 = 0–2.99 MET-hours/week; 2 = 3–5.99 MET-hours/week; 3 = ≥6 MET-hours/week). Six or more MET hours per week are equivalent to >30 minutes of moderate-intensity leisure time physical activity 3 times per week, which is considered being physically active for persons with chronic illnesses (28, 29).
A Pearson's correlation matrix was computed with all the variables to examine the bivariate correlations between the outcome variable (sleep quality) and each potential predictor variable. The pattern of intercorrelations among the possible predictor variables was also examined.
A series of hierarchical multiple regression analyses were computed to test the importance of disease-related, exercise, and psychosocial factors (tested in sets) to sleep quality after controlling for demographic factors. Hierarchical multiple regression is the regression strategy of choice when the research goals are to determine the importance of a predictor variable(s) once other predictor variables have already been entered into the equation (30). Each hierarchical regression analysis determined whether the variance explained by the specific set (i.e., disease-related, exercise, or psychosocial) contributed significantly to the total variance in sleep quality, after controlling for the demographic set. To this end, as suggested by Cohen and Cohen (30) the increment in R2 was tested for statistical significance. A standard multiple regression analysis was then computed with all the variables entered simultaneously into the model to determine the relative contribution of these variables to sleep quality. Variable selection was based on theoretical relevance, pattern of correlation with the outcome variable and other potential predictor variables, and the assumptions underlying multiple regression analysis.
Of the 106 women with SLE who agreed to participate in this study, 100 (94.3%) returned their completed self-report questionnaires. Of the 6 patients who failed to return their questionnaires, 4 were no longer interested in participating and 2 felt too ill to return the questionnaires.
Table 1 summarizes the demographic and clinical characteristics for the 100 patients. The mean ± SD age was 45.19 ± 14.12 years. Seventy-eight percent (78%) of the women were Caucasian. The mean ± SD SLAM-R score was 5.56 ± 3.52 indicating mild to moderate disease activity; the mean ± SD SLICC/ACR DI score was 1.69 ± 1.81 (interquartile range 3.0). Twenty-six patients (26%) reported taking prednisone in the last 3 months. Of these 26 patients, 11 (42%) were taking ≤7.5 mg/day, and 15 (58%) were taking >7.5 mg/day. Fifteen (15%) patients fulfilled the ACR criteria (31) for fibromyalgia. The mean ± SD score on the CESD-AR was 13.02 ± 11.35, with 29 patients (29%) scoring at or above the standard cutoff score (≥16) for clinically significant levels of depression on the CES-D (9). Forty-six percent of the sample were exercising regularly (exerting 6 or more metabolic equivalent hours a week).
|Mean ± SD||Median|
|Age, years||45.19 ± 14.12|
|Education, years||13.17 ± 3.06|
|Personal income†||4.28 ± 1.78||5.00|
|Smokers, n (%)||18 (18)|
|SLAM-R||5.56 ± 3.52|
|SLICC/ACR DI||1.69 ± 1.81||1.00|
|Disease duration, years||13.14 ± 9.43|
|HAQ index||0.37 ± 0.52|
|FM diagnosis, n (%)||15 (15)|
Table 2 shows the global score and the component scale scores for the PSQI for our sample. Published comparative data for healthy controls and primary insomniacs are also presented (6). To facilitate comparison of the PSQI scores obtained in our sample to these published norms, one-sample t-tests were computed. This analysis allows single samples to be compared with standardized norms. The mean ± SD global PSQI score for the SLE patients was 6.98 ± 4.03 (range 1–18), with 56 patients (56%) classified as “poor sleepers” (global PSQI ≥6). SLE patients had significantly poorer scores on all the component scores and global score compared with healthy patients. Severity of problems with sleep latency (includes items related to how many minutes needed to fall asleep and the frequency of inability to fall asleep within 30 minutes) and sleep disturbance was comparable with levels reported by patients with primary insomnia. Daytime dysfunction (which includes items related to daytime sleepiness and energy) was significantly worse for SLE patients compared with patients with primary insomnia.
|SLE (n = 100)||Healthy controls† (n = 52)||Insomniacs† (n = 45)|
|Global score||6.98 ± 4.03‡||2.67 ± 1.70||10.38 ± 4.57|
|Sleep quality||1.21 ± 0.80‡||0.35 ± 0.48||1.96 ± 0.93|
|Sleep latency||1.22 ± 1.15§||0.56 ± 0.73||1.42 ± 1.01|
|Sleep duration||0.93 ± 1.00‡||0.29 ± 0.50||1.51 ± 1.20|
|Sleep efficiency||0.63 ± 1.00‡||0.10 ± 0.30||1.47 ± 1.24|
|Sleep disturbance||1.51 ± 0.67§||1.00 ± 0.40||1.40 ± 0.62|
|Use of sleep medications||0.45 ± 0.98‡||0.04 ± 0.28||1.20 ± 1.31|
|Daytime dysfunction||1.15 ± 0.83‡||0.35 ± 0.48||1.42 ± 0.94|
Pearson coefficients were computed to identify correlates of sleep quality in SLE patients. As shown in Table 3, cumulative damage, disease activity, functional disability, pain severity, use of prednisone, and depressed mood were positively correlated with poorer sleep quality. Lower levels of exercise participation were significantly associated with poorer sleep quality.
|SLICC/ACR DI||0.22||< 0.05|
|HAQ index||0.41||< 0.01|
|HAQ pain||0.18||< 0.10|
|Use of prednisone||0.28||< 0.01|
|Level of exercise participation||−0.38||< 0.001|
The results of the hierarchical multiple regression analyses are shown in Table 4. The first model testing the contribution of demographic variables to sleep quality was not found to be statistically significant (F[2,97] = 1.92, R2 = 0.04, P = 0.15). The addition of disease-related variables (model 2) resulted in a significant increase in the R2 value (F[8,91] = 4.04, R2 = 0.26, P < 0.001). Use of prednisone and worse functional disability contributed to poorer sleep quality (P = 0.006 and P = 0.003, respectively). In Model 3, the addition of level of exercise participation significantly added to the demographic set (F[3,96] = 6.41, R2 = 0.17, P = 0.001). The addition of depressed mood also significantly added to the demographic set (F[3,96] = 14.16, R2 = 0.31, P < 0.001), explaining an additional 29% of the variance in sleep quality. In this model higher depressed mood scores were significantly associated with poorer sleep quality (P < 0.001).
|Model 1 Demographic||Model 2 Disease related||Model 3 Exercise||Model 4 Psychosocial||Full Model|
The results of the standard multiple regression model in which all the variables were entered simultaneously are also presented in Table 4. Only depressed mood remained a significant determinant of sleep quality (P < 0.001), although a trend was observed for use of prednisone (P = 0.055) and lower level of exercise participation (P = 0.088) with poorer sleep quality.
In the present study, the prevalence of poor sleep in patients with SLE was 56%. Tench et al (32) reported a similar rate (60%) using the PSQI in their sample of SLE patients. Other studies show that the prevalence of sleep disturbance in SLE appears to be higher than those shown in healthy controls and in the general population (range 9–40%) (1, 2, 6).
We found support for the biopsychosocial model in understanding sleep quality in women with SLE. The most important determinant of sleep quality was depressed mood. This association remained strong, even after controlling for demographic factors, disease-related variables, and level of exercise participation. The multivariate models also found prednisone use and engaging in less exercise to be associated with poorer sleep quality.
A relationship between depressed mood and respiratory disturbances during sleep in SLE patients using polysomnographic data has previously been demonstrated (4). The relationship between sleep disturbances and depressed mood has also been shown in epidemiologic studies of sleep in the general population (33–35), and with other medical patients (i.e., arthritis, chronic pain) (36–38). Although the cross-sectional nature of the present study precludes making causal inferences regarding sleep disturbance and depression, other studies suggest a bidirectional relationship. Epidemiologic and electroencephalographic sleep studies have shown a role for sleep disturbances in the pathogenesis of depression (39–43). However, there are longitudinal data to suggest that increases in depressive symptoms worsens sleep quality (44).
Exercise participation was associated with better sleep quality. To our knowledge, this is the first study to report this relationship in patients with SLE. The potentially beneficial effect of exercise on sleep has only recently been the subject of investigation. Our findings are consistent with studies demonstrating a useful role of exercise on sleep quality in healthy populations (45, 46) and extend the relationship to women with SLE. The mechanisms by which exercise may improve sleep quality require further study. There is some evidence to suggest physiologic pathways including muscular relaxation, decreases in sympathetic tone, or thermal changes induced by exercise may promote sleep (47–49). Exercise has also been associated with improvements in depressed mood and anxiety levels (50–52) that influence sleep quality. Exercise participation was associated with less depressive symptoms in our study. While we acknowledge that this association was cross-sectional, it suggests that this psychosocial mechanism is worthy of further study. The potentially beneficial effect of exercise has important clinical implications, because moderate-intensity exercise has been shown to be a useful nonpharmacologic intervention for enhancing sleep quality (53) and psychological well being (50–52).
Steroid therapy can contribute to sleep difficulties. The use of prednisone in the present study was associated with sleep disturbances, although other studies have not shown a relationship between use of corticosteroids and sleep disturbance in SLE (4, 5, 32). Although fewer patients in our sample were receiving corticosteroids, a greater proportion were taking higher doses of corticosteroids compared with previous studies (4, 5, 32), which may partially explain our results.
Bivariate analyses indicated a significant relationship between disease activity and sleep disturbances. However, this association did not remain significant in the multiple regression models, suggesting that the relationship between sleep and disease activity may be weak and nonlinear. Alternatively, disease activity may indirectly influence sleep quality. That is, patients in a more active state are more likely to be taking prednisone and less likely to engage in regular physical activity. The bivariate relationship between pain severity and sleep disturbances was not apparent in the multiple regression models. This may be because pain intensity was on average low and perhaps not severe enough to disrupt sleep. In their sample of patients with nonmalignant chronic pain, Menefee et al (54) have shown that pain interferes with sleep quality only in patients reporting higher levels of pain intensity. However, other studies have also shown weak or no relationship between pain intensity and sleep complaints in various medical populations (32, 38, 55).
The present study has several limitations. Sleep quality was assessed by self report. Sleep disturbances are assessed most accurately with the use of polysomnography. We did, however, select a sleep measure that has previously been validated and compared with polysomnography. In our final model, 32% of the variance in sleep quality was explained, suggesting that other variables not assessed in our study contribute to sleep quality in patients with SLE. The cross-sectional design of the present study does not allow us to determine the direction of the relationships found. Future multivariate prospective studies are needed to expand our understanding of sleep disturbances among SLE patients. Finally, our patient sample was fairly well-educated, consisting predominately of middle-class Caucasian women drawn from a tertiary care lupus clinic, limiting generalizability. Despite these limitations, our study remains one of the best efforts to date in assessing sleep quality in SLE.
In conclusion, the majority of SLE patients exhibit sleep disturbances. Poor sleep may play a more important role than is generally believed in the fatigue experienced in patients with SLE. Our findings highlight the need to routinely assess sleep quality in this patient population. The PSQI (a brief, self-administered instrument) could be used in a clinical setting to measure sleep complaints in SLE patients. Modifiable determinants of sleep quality (such as depressed mood and lack of regular exercise) may be important areas to target in interventions aimed at promoting sleep in patients with SLE. This in turn might improve the fatigue that greatly affects women with SLE (32).