Effects of a stress-reduction program on psychological function, pain, and physical function of systemic lupus erythematosus patients: A randomized controlled trial




To assess the effects of a stress-reduction program on pain, psychological function, and physical function in persons with systemic lupus erythematosus (SLE) who experience pain.


Ninety-two SLE patients were assigned randomly to receive either biofeedback-assisted cognitive-behavioral treatment (BF/CBT), a symptom-monitoring support (SMS) intervention, or usual medical care (UC) alone.


BF/CBT participants had significantly greater reductions in pain and psychological dysfunction compared with the SMS group (pain, P = 0.044; psychological functioning, P < 0.001) and the UC group (pain, P = 0.028; psychological functioning, P < 0.001). BF/CBT had significantly greater improvement in perceived physical function compared with UC (P = 0.035), and improvement relative to SMS was marginally significant (P = 0.097). At a 9-month followup evaluation, BF/CBT continued to exhibit relative benefit compared with UC in psychological functioning (P = 0.023).


This study supports the utility of a brief stress management program for short-term improvement in pain, psychological function, and perceived physical function among persons with SLE who experience pain.


Although medical management of systemic lupus erythematosus (SLE) has resulted in decreased mortality, the disease continues to impact quality of life. Pain, fatigue, changes in social roles, and the unpredictability of flares create an additional burden for many patients. Pain affects up to 90% of patients. Psychological distress, widespread in SLE (1, 2), has been linked to increased SLE disease activity (3, 4) and reductions in work productivity (5). Patients with SLE may be reluctant to add further analgesics, opioids, and antidepressants to the numerous agents necessary for basic management of SLE. Therefore, nonpharmacologic methods may be advantageous. Behavioral methods, such as biofeedback, relaxation, and cognitive-behavioral therapy, have led to decreased pain and distress and improved function in persons with chronic pain of various etiologies, including rheumatoid arthritis and fibromyalgia (6–9). The aim of the present investigation was to evaluate the utility of a biofeedback-assisted cognitive-behavioral treatment (BF/CBT) program for persons with pain related to SLE.

A limited number of studies have evaluated psychological interventions to improve the quality of life for persons with SLE. Braden and colleagues (10) described a positive impact of SLE self-help groups on depression, self worth, and self efficacy for managing SLE symptoms. Although the results described are encouraging, no control group was used and the outcome measures were not standardized, so it is difficult to draw definitive conclusions from this research. Austin et al (11) evaluated the effects of a person-centered telephone counseling program designed to enhance coping with SLE. The counseling intervention was associated with improvements in perceived physical function compared with a telephone symptom-monitoring control group. No differences were found between the groups on pain or psychological functioning. A recent randomized, controlled multisite study of supportive-expressive group psychotherapy found that this intervention was not associated with improvements in psychological distress, perceived health status, health care utilization, or disease activity compared with a standard care control group (12). However, by the 6- and 12-month followup evaluations, the group therapy participants exhibited reductions in illness intrusiveness, or the extent to which SLE and its treatment interfered with intimacy, social relationships, and personal development (13). Pain was not a specific focus of this group, rather the goals were to foster social support, address emotional issues, and facilitate coping skills. No controlled studies of BF/CBT for pain and stress related to SLE have been reported previously.

The primary purpose of this study was to evaluate the effects of a BF/CBT program on physical functioning, pain, and psychological adaptation of persons with SLE. Exploring the impact of the intervention on SLE disease status was considered a secondary aim, due to the fluctuating nature of SLE disease activity. The 6-session treatment program was administered individually rather than in a group or via telephone. This intervention, although brief, included intensive face-to-face contact with the treating professional. Such contact could result in improvement simply due to contact rather than the specific active ingredients of relaxation and problem-solving skills. Therefore, 2 conditions were included for comparison. A 6-session symptom-monitoring support (SMS) condition was designed to control for the potential nonspecific effects of time and attention from the therapist. The other control condition included standard or usual medical care (UC) only. We hypothesized that persons with SLE treated with BF/CBT would exhibit greater reductions in pain and psychological distress and greater improvement in physical functioning than the UC control group and the SMS attention-control group at posttreatment and followup assessments.



Participants were recruited through a mailing to the Pittsburgh Lupus Registry, which contains >900 living individuals with SLE. Subjects meeting American College of Rheumatology 1982 revised criteria for SLE (14) whose medication regimen had been stable for at least 1 month and those reporting at least a 3-month history of pain, with pain currently occurring at least 3 times each week, were included. Persons with active renal involvement and those taking >15 mg prednisone or the equivalent on a daily basis were excluded. These exclusion criteria were necessary to avoid potential confounding of treatment effects with medication effects.

Of 210 persons responding to the recruitment mailing, 29 (13.8%) reported that they currently were not experiencing pain and 2 (<1%) were taking >15 mg prednisone or the equivalent and therefore were not included. An additional 70 persons were not interested in participating in the study due to lack of time, transportation, and distance to the facility (10), family reasons (7), or did not state a reason (53). Those patients (n = 109) who were eligible upon initial screening were invited to attend the initial study evaluation visit. Of the 109, 93 persons kept their clinic appointment. One of these was not eligible for medical reasons; therefore, 92 were eligible at the initial evaluation visit.


Prior to study entry, participants completed informed consent documents approved by the University of Pittsburgh Institutional Review Board. At the initial visit, participants received a medical evaluation of current lupus activity conducted by a rheumatologist specializing in SLE (SM) and completed psychosocial questionnaires (described below). At the end of the initial visit, participants who qualified for inclusion were assigned randomly, based on a software-generated randomization plan, to receive either a biofeedback-assisted cognitive-behavioral pain and stress management treatment program, symptom monitoring and support, or usual medical care alone. Following the 6-session intervention phase, participants repeated the psychosocial questionnaires. At 3 and 9 months postintervention, participants received followup medical evaluations. Questionnaires were repeated at the 9-month followup evaluation. Those assigned to UC completed postintervention and followup evaluations on the same schedule as those assigned to BF/CBT and SMS. Potential bias regarding treatment outcome was minimized because the rheumatologist who completed the medical evaluations and the research associate who administered the self-report questionnaires were masked to treatment assignment. Participants were instructed not to discuss their treatment condition with the rheumatologist or research associate.


Multiple measures were chosen to represent important conceptual domains relevant to living with lupus. The domains of impact of lupus and potential treatment outcomes are described below, along with the selected measurement instruments.


Two psychometrically validated self-report measures of pain were included. The Revised Arthritis Impact Measurement Scales, pain subscale (AIMS2-Pain) (15) consists of 5 averaged items (score range 0–4) that assess frequency and severity of pain and stiffness. The pain interference scale of the Multidimensional Pain Inventory (MPI-I) (16) is an 11-item scale assessing the impact of pain on the patient's family, social activities, and work-related activities during the previous week. Each item is rated on a 0–6 scale, and the completed items are averaged.

Psychological functioning.

Psychological functioning was assessed by 3 validated measures. The Center for Epidemiological Studies Depression scale (CES-D) (17) is a 20-item inventory that assesses presence and frequency of depressive symptoms over the past week, yielding a score range of 0–60. Cohen's Perceived Stress Scale, 4-item version (STRESS) (18) assesses the frequency of feeling out of control or overwhelmed with difficulties during the past week, and can range from 0 (none) to 16 (high stress). The Arthritis Self-Efficacy Scales, pain and other symptoms subscale (19), consists of 11 items designed to measure confidence in one's ability to manage the pain, fatigue, frustration, and other aspects of disease; it was reworded to reflect lupus rather than arthritis (3). Confidence estimates for each item range from 0% to 100% and competed items are averaged, yielding a score range of 0–100.

Perceived physical functioning.

Participants' perception of their physical functioning was assessed by 3 measures. The Short Form 36 Health Survey (SF-36) (20) is a reliable and valid instrument that frequently is included in quality of life studies of persons with lupus (4, 21). The physical function scale of the SF-36 (SF-36-PF) was used in the present study. The SF-36-PF consists of 10 questions regarding perceived limitations in various activities. Scores may range from 0 to 100, with higher scores indicating better functioning. The 9-item Fatigue Severity Scale (22) assesses the impact of fatigue on various activities, such as physical functioning and carrying out responsibilities. Items are scaled from 0 to 6, with higher scores indicating greater severity, and are averaged. The patient global assessment of functioning is a 10-cm visual analog scale on which the patient marks how well they are doing based on all the ways that SLE affects them using verbal anchors of “very well” (0 cm) and “very poorly” (10 cm). This scale has been demonstrated to significantly correlate with self-reported physical disability (r = 0.56) in lupus patients (23) and physician judgment of SLE activity (2).

SLE disease activity.

The Systemic Lupus Erythematosus Disease Activity Index (SLEDAI) (24) contains 24 weighted descriptors of signs and symptoms present in the past 10 days, and scores can range from 0 (no activity) to 105. The Systemic Lupus Activity Measure, Revised (SLAM-R) (25) measures disease activity in 11 organ systems over the previous month. Scores on the SLAM-R can range from 0 to 81, and a score of ≥7 is considered a level at which most expert rheumatologists would initiate treatment (26).



Participants received a standardized 6-session protocol that included auditory electromyographic biofeedback from the trapezius area, progressive muscle relaxation, and cognitive-behavioral pain and stress management training. Cognitive-behavioral training included problem-solving strategies for coping with pain, other lupus symptoms, and interpersonal issues; the training also included recognition and alteration of automatic maladaptive thinking patterns that can contribute to stress and pain. This protocol was administered by a licensed psychologist who had been conducting research based on this protocol for 6 years. The intervention included homework assignments to practice and record use of the relaxation techniques and to complete worksheets regarding stressor identification and problem-solving skills.


This condition was designed to control for nonspecific effects, such as potential benefits of receiving attention from the therapist. During the 6 sessions, participants were asked to describe their lupus symptoms over the previous week. They were invited to report on their lupus history, family history of autoimmune disease, and any current or past stressors. The therapist's role was to elicit information and listen empathically but avoid giving suggestions for change or teaching any pain or stress-management skills. This protocol was administered by the same provider as the BF/CBT intervention. Homework included monitoring of SLE symptoms on a daily basis.


Persons assigned to the UC condition did not receive any study interventions, but completed followup evaluations on the same schedule as BF/CBT and SMS participants.

All participants in each of the study conditions continued with their usual medical care (such as taking nonsteroidal antiinflammatory drugs, oral prednisone, and hydrochloroquine) without restrictions.

Treatment integrity and adherence.

To ensure treatment integrity, BF/CBT and SMS sessions followed written protocols and the therapist completed a standardized checklist for each session. All BF/CBT and SMS sessions were audiotaped and the therapist's supervisor independently rated which treatment, BF/CBT or SMS, had occurred in a random selection of 10 tapes. Participants in the BF/CBT and SMS groups rated the credibility of their intervention and expectancy of improvement on a 5-item scale (27) following the first treatment session. For this scale, questions such as, “How logical does the treatment that you are receiving seem to you?” and “How confident are you that this treatment will be successful in helping you with your lupus pain?” are rated from 0 (illogical, not confident) to 6 (very logical, very confident) and averaged. Persons in the BF/CBT condition kept diaries of relaxation practice during the intervention period. Additionally, they rated the frequency of relaxation practice on a 0–3 scale (0 = never, 1 = rarely, 2 = sometimes, 3 = very often) at 3 and 9 months posttreatment. The number of diaries returned and self ratings of relaxation practice frequency were used to estimate treatment adherence.

Data analysis.

The basic experimental design and corresponding data analyses represented a one-between subjects (BF/CBT versus SMS versus UC) and one-within or repeated factor (pre, post, and followup assessments) crossed factorial design, with multiple dependent outcome measures. A multivariate or doubly multivariate (multiple qualitatively distinct measures at multiple times) approach to repeated measures (28, 29), based on the unweighted general linear model, was used to test for pre–post–followup changes between the treatment group (BF/CBT) and the 2 control groups (SMS and UC). This approach provides better control for experimentwise error rates and also avoids the compound symmetry assumption necessary for the univariate repeated measures approach (30). This approach also adjusts statistically for any initial differences between the 3 groups, and is easier to successfully implement and generally more powerful than the use of covariant models when multiple outcome measures are considered simultaneously (31). Primary analyses were based on a priori multivariate contrasts designed to test our primary hypotheses. Significant interactions were interpreted by simple main effects; the Tukey-Kramer method (32) of controlling Type I error rates was used for between-subject post-hoc comparisons; and improved, more powerful Bonferroni-type multiple comparison procedures (30) were used to control Type I error rates for within-subject comparisons. Prior to conducting the primary statistical analyses, we evaluated the statistical models for violations of homogeneity of errors and nonlinearity using standard graphic methods. Because of the smaller sample size, exact probability chi-square analyses, computed with the StatXact program (33) were used to analyze dichotomous and ordinal measures. P ≤ 0.05 was used to indicate statistical significance.

The data from all randomized participants were analyzed using an intent-to-treat model. Multiple imputation (34) was used to estimate missing data from dropouts and those participants who did not complete all followup evaluations. Multiple imputation, first proposed in the early 1970s (35), uses ordinary least-squares regression methods to compute multiple imputations based on actual values, and replaces each missing value in the data set with multiple imputed values instead of just one, thus reflecting the uncertainty surrounding the value to be imputed (36–38).

Statistical power calculations in the original planning for this trial indicated that an initial sample of 120 subjects (40/group) would result in an estimated power of 0.84 to detect 9-month moderate effect size differences (0.5 SD) between the active treatment (BF/CBT) and attention control (SMS) groups, and >0.95 power at 9 months to detect major effect size (0.8 SD) differences between the active treatment and usual care control (UC) conditions. Due to recruitment difficulties, as well as time and funding considerations, the originally planned sample sizes could not be obtained. Nonetheless, we believe that the resulting sample sizes were still adequate for testing our primary hypotheses. Our revised power estimates, based on the obtained sample sizes, indicated an acceptable, although not ideal, power of 0.77 to detect moderate effect size differences between the groups at the time of the 9-month followup evaluation.


Subject assignments.

As displayed in Figure 1, 92 persons were determined to be eligible, completed informed consent procedures, and were randomly assigned to receive BF/CBT (n= 32), SMS (n= 33), or UC (n= 27). Seven persons (3 BF/CBT, 4 SMS) declined participation prior to the intervention phase but after randomization, primarily due to lack of time. Four persons (2 each from the BF/CBT and SC/SM groups) dropped out of the study after attending 1 or 2 of the 6 intervention sessions. Reasons cited included time constraints (1 BF/CBT, 1 SMS) or dissatisfaction with the assigned treatment (1 BF/CBT, 1 SMS). Two persons assigned to UC did not complete all followup evaluations. Therefore, the total attrition rate after randomization was 14%, although 5 of these 13 patients (3 BF/CBT, 1 SMS, and 1 UC) returned to complete a followup medical and psychosocial evaluation. Analysis of variance (ANOVA) and chi-square analyses indicated that those who declined participation after randomization or dropped out of the study did not differ significantly from study completers with regard to demographic characteristics or dependent variables at pretreatment.

Figure 1.

Trial profile. BF/CBT = biofeedback-assisted cognitive-behavioral treatment; SMS = symptom-monitoring support; UC = usual medical care.

Demographic measures.

The baseline demographic characteristics of the participants are presents in Table 1. Participants' ages ranged from 23 to 69 years, and disease duration ranged from <1 year to 36 years. Cumulative damage due to SLE, as assessed on the Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index, ranged from 0 to 8, with a mean ± SD of 1.38 ± 1.71. Eighty-seven of the participants were women. Seventy-one (77%) were white, 20 were African American, and 1 (assigned to SMS) was Asian. ANOVA and chi-square analyses indicated that the 3 groups did not differ significantly on any of the demographic characteristics (Table 1). The sample was representative of the Pittsburgh Lupus Registry in terms of age and race. Disease duration of this sample (10.99 years on average) was shorter than that of the entire registry (13.7 years; P < 0.001).

Table 1. Participant demographics, by group*
 BF/CBT (n = 32)SMS (n = 33)UC (n = 27)P
  • *

    BF/CBT = biofeed back-assisted cognitive-behavioral stress management; SMS = Symptom-monitoring support; UC = usual medical care; SLICC/ACR = Systemic Lupus International Collaborating Clinics/American College of Rheumatology.

Age, mean ± SD years48.2 ± 9.146.7 ± 11.747.0 ± 10.50.85
Disease duration, mean ± SD years12.0 ± 7.711.2 ± 7.19.6 ± 6.00.41
SLICC/ACR Damage Index, Mean ± SD1.7 ± 1.81.4 ± 1.60.9 ± 1.70.21
Female, %9497930.74
Ethnicity, %    
 African American28.118.218.5 

Treatment credibility and adherence measures.

The BF/CBT and SMS interventions appeared to be equally credible to participants. Mean ± SD treatment credibility ratings were 5.11 ± 0.99 and 5.18 ± 0.81 (P = 0.76) for BF/CBT and SMS, respectively. The supervising psychologist correctly identified the treatment condition in all 10 randomly selected audiotapes. Twenty-five of the 27 BF/CBT participants turned in at least 3 weekly relaxation practice diaries. Of these participants, 96% recorded practicing 5 or more days/week. Relaxation practice was rated as “very often” by 41% of participants 3 months posttreatment and 48% at the 9-month followup assessment, whereas 50% and 44% practiced “sometimes” at 3-month and 9-month followup evaluations, respectively.


Our primary hypothesis was that the active treatment, BF/CBT, would lead to greater improvements in pain, psychological functioning, and physical functioning following treatment than each of the 2 control groups (SMS and UC). A secondary aim was to explore whether the active treatment was associated with greater reductions in SLE disease activity compared with the control groups.

Short-term effects of BF/CBT.

A multivariate ANOVA (MANOVA) approach to repeated measures, described above, was used to compare the BF/CBT outcome data (pre- to posttreatment) with each of the control conditions in the areas of pain, psychological functioning, and physical functioning. Means, standard deviations, effect sizes, and P values for the primary hypotheses by outcome domain are displayed in Table 2.

Table 2. Means ± SDs by intervention group for outcome measures at baseline, posttreatment, and followup, and effect sizes and P values for planned hypotheses*
Outcome measureGroupBaselinePosttreatment9-month followupEffect sizesP values for hypotheses
Pre–postPre–9 mo.Pre–Post BF/CBT > UCPre–post BF/CBT > SMSPre–9 mo. BF/CBT > UCPre–9 mo. BF/CBT > SMS
  • *

    mo. = months; AIMS2-Pain = Revised Arthritis Impact Measurement Scales, pain subscale; MPI-I = Multidimensional Pain Inventory, pain interference scale; SF-36 PF = Short Form 36 Health Survey, Physical Function scale. For other abbreviations, see Table 1 footnote.

  • BF/CBT n = 32, SMS n = 33, UC n = 27.

  • Significance levels for these items were determined with doubly multivariate analysis of variance (multiple measures at multiple times).

Pain      0.0280.0440.3050.718
 AIMS2-PainBF/CBT2.58 ± 0.741.98 ± 0.872.05 ± 0.940.740.630.0880.092  
 SMS2.30 ± 0.831.97 ± 0.911.87 ± 0.950.370.47    
 UC1.96 ± 0.811.65 ± 0.891.69 ± 1.150.360.27    
 MPI-IBF/CBT3.81 ± 1.302.86 ± 1.363.04 ± 1.340.720.590.0080.014  
 SMS3.21 ± 1.532.83 ± 1.612.67 ± 1.570.250.35    
 UC2.76 ± 1.552.46 ± 1.632.43 ± 1.970.190.19    
Psychological function      < 0.0010.0010.0230.215
 Depressive symptomsBF/CBT20.31 ± 11.8214.86 ± 10.0715.01 ± 10.500.500.480.0120.1040.267 
 SMS17.82 ± 11.2816.52 ± 11.5314.17 ± 11.640.110.32    
 UC18.96 ± 12.3420.33 ± 14.1416.99 ± 12.94−0.100.18    
 Self-efficacyBF/CBT50.46 ± 20.7471.21 ± 18.9864.40 ±< 0.001< 0.0010.008 
 SMS53.50 ± 22.3954.25 ± 24.1055.74 ± 20.960.030.10    
 UC61.71 ± 19.0458.36 ± 20.3260.04 ± 23.72−0.17−0.08    
 Perceived stressBF/CBT6.97 ± 3.655.32 ± 3.075.62 ± 2.920.490.410.0060.0210.048 
 SMS5.88 ± 3.156.07 ± 3.514.84 ± 3.26−0.060.32    
 UC6.07 ± 3.406.73 ± 4.046.32 ± 3.38−0.18−0.07    
Physical function      0.0350.0970.1250.227
 SF-36-PFBF/CBT42.34 ± 23.5552.18 ± 22.8850.48 ± 25.530.420.330.106   
 SMS48.49 ± 25.3649.13 ± 26.7250.22 ± 23.860.030.07    
 UC58.33 ± 23.9860.98 ± 25.6356.95 ± 27.300.11−0.05    
 FatigueBF/CBT4.79 ± 1.154.29 ± 1.684.58 ± 1.690.360.150.139   
 SMS4.74 ± 1.094.45 ± 1.504.34 ± 1.460.280.31    
 UC4.13 ± 1.434.11 ± 1.773.93 ± 1.740.090.13    
 Global self- assessmentBF/CBT5.13 ± 2.353.59 ± 2.723.68 ± 2.510.610.600.009   
 SMS4.36 ± 2.474.18 ± 2.443.75 ± 2.500.070.24    
 UC4.02 ± 2.554.63 ± 2.823.67 ± 2.58−0.230.14    


The results of the MANOVA for the 2 pain measures, AIMS2-Pain and MPI-I, are presented in Table 2. Changes in pain measures were significantly greater for the BF/CBT group compared with both the UC group (P = 0.028) and the SMS group (P = 0.044). Analysis of simple main effects using MANOVA contrasts indicated that the BF/CBT group displayed statistically significant reductions in the pain measures from pre- to posttreatment (P = 0.011), but the SMS (P = 0.452) and UC (P = 0.224) groups did not.

Psychological functioning.

Table 2 presents results for the measures that were used to assess changes in psychological functioning: CES-D, Arthritis Self-Efficacy Scale, and the STRESS. The BF/CBT group exhibited significantly greater improvement in these measures from pre- to posttreatment compared with both the UC group (P < 0.001) and the SMS group (P = 0.001). Simple main effects indicated that the psychological functioning measures improved only in the BF/CBT group (P < 0.001), did not change significantly in the SMS group (P = 0.225), and actually worsened from pre- to posttreatment for the UC group (P = 0.016).

Physical functioning.

Patients' perceptions of their general physical functioning was assessed by 3 measures that included fatigue, physical function limitations, and global evaluation of overall lupus impact. As displayed in Table 2, the BF/CBT group had significantly greater improvement compared with the UC group (P = 0.035); however, their improvement relative to the SMS participants was only marginally significant (P = 0.097). Analysis of simple main effects indicated a significant improvement on the physical functioning measures for the BF/CBT group (P = 0.025), but no significant pre- to posttreatment changes for SMS (P = 0.475) or UC (P = 0.158).

Disease activity.

Table 3 provides information regarding disease activity in the 3 groups at pretreatment and the 3-month followup. The disease activity changes observed for the BF/CBT group were not significantly different from those found for the SMS group (P = 0.220) or the UC group (P = 0.372). A MANOVA for the time main effect was found to be significant (P < 0.001), indicating that the sample as a whole showed a statistically significant reduction in disease activity between the pretreatment and 3-month followup evaluations. As can be seen from the effect sizes displayed in Table 3, the SLAM-R measure displayed larger changes then did the SLEDAI. Analysis of simple main effects using MANOVA contrasts indicated that none of the groups individually displayed significant changes for the disease activity measures between the pretreatment and 3-month followup assessments.

Table 3. Means ± SDs and effect sizes by intervention group for disease activity measures at baseline, and 3- and 9-month followup evaluations*
Outcome measureGroupBaseline3-month9-monthEffect sizes
Pre–3 mo.Pre–9 mo.
  • *

    None of the multivariate comparisons between groups across time were statistically significant. mo. = months; SLAM-R = Systemic Lupus Activity Measure, Revised; SLEDAI = Systemic Lupus Erythematosus Disease Activity Index. For other abbreviations, see Table 1 footnote.

  • BF/CBT n = 32, SMS n = 33, UC n = 27.

Disease activity      
 SLAM-RBF/CBT9.72 ± 3.807.44 ± 2.527.95 ± 3.140.720.57
 SMS8.27 ± 3.436.73 ± 3.086.31 ± 2.720.470.64
 UC6.78 ± 2.615.58 ± 2.355.76 ± 3.370.480.34
 SLEDAIBF/CBT3.91 ± 3.133.31 ± 2.942.85 ± 2.380.200.39
 SMS4.03 ± 3.752.78 ± 2.152.87 ± 2.210.430.39
 UC2.56 ± 2.172.20 ± 2.192.30 ± 2.720.160.11

Longer-term effects of BF/CBT.

Similar to the analyses above, a MANOVA approach to repeated measures was used to evaluate longer-term effects of BF/CBT, that is, changes from pretreatment to the 9-month followup visits, and to compare these outcomes to the 9-month followup data observed for the 2 control groups.


As displayed in Table 2, although the BF/CBT group displayed the largest pretreatment–9-month effect sizes for the pain domain, the significant advantage of BF/CBT for pain reduction observed for the short-term outcome was not maintained by the 9-month followup evaluation. A MANOVA of simple main effects indicated that the values of the pain measures at 9 months for the BF/CBT group were not significantly different from their pretreatment values (P = 0.392).

Psychological functioning.

As displayed in Table 2, the BF/CBT group had significantly greater improvement for these measures at the 9-month followup compared with the UC group (P = 0.023), but not when compared with the SMS group (P = 0.215). However, simple main effects indicated that compared with pretreatment psychological functioning measures, the 9-month followup measures for the BF/CBT remained significantly improved (P = 0.036). No significant changes were observed for the SMS group (P = 0.658), and the psychological functioning measures remained marginally worse for the UC group (P = 0.072). As can be seen in Table 2, self efficacy for lupus symptom management and perceived stress were the specific psychological measures that contributed to the BF/CBT participants' maintenance of improvement relative to the other 2 groups by the 9-month followup. In contrast, the CES-D did not appear to be a sensitive measure of differences between these groups.

Physical functioning.

Table 2 also shows that the short-term superiority of the BF/CBT treatment for measures of physical functioning was not maintained at the 9-month assessment. Simple main effects indicated that these measures for the BF/CBT group at 9 months were only marginally better than their pretreatment values (P = 0.086).

Disease activity.

No significant differences between the groups were found for disease activity measures. However, similar to the short-term effects, a MANOVA for the pretreatment–9-month time effect was significant (P < 0.001), indicating that the sample as a whole showed statistically significant improvement in disease activity between the pretreatment and 9-month followup assessments.

Effect size estimates for each of the outcome variables are provided in Tables 2 and 3. We consider a moderate effect size of at least 0.5 to indicate a clinically important improvement (39). Moderate to large effect sizes were found for BF/CBT in the areas of pain severity and pain interference, self efficacy for managing symptoms, global evaluation of functioning, SLAM-R, and depressive symptoms from pre- to posttreatment. Similar effect sizes were observed for pretreatment to 9-month outcomes for all of these but depressive symptoms. Effect sizes were <0.5 for the control conditions for all variables with the exception of the SLAM-R pretreatment to 9-month effect size of 0.64 in the SMS group.

The analyses reported above were repeated using data from only those participants who completed the entire program. The same pattern of results was observed in this efficacy analysis as was found in the intent-to-treat analyses, described above.


The results of the study indicate that a brief BF/CBT intervention for SLE leads to greater short-term improvement in pain, psychological function, and perceived physical function compared with symptom monitoring and usual care. The BF/CBT program appeared to enhance the participants' perceptions of efficacy for managing symptoms, maintaining valued activities, and general well-being, as supported by relatively large pretreatment–posttreatment effect sizes in self efficacy for managing symptoms, pain interference, and global assessment of function. The improvement in psychological functioning relative to usual care was maintained at a 9-month followup assessment.

Two control groups were included to determine whether BF/CBT may be a useful behavioral intervention beyond usual medical care, and to assess whether BF/CBT has specific effects beyond those of an attention-placebo condition. The SMS condition was designed to control for any nonspecific effects of participation, such as receiving time and attention from the therapist. However, it is difficult to devise a credible attention-placebo intervention that has no treatment effects. Symptom monitoring has been used as a comparison condition in a previous telephone counseling intervention study (40) and, consistent with our findings, was associated with small to medium effect sizes for fatigue, affect, and pain. In the current study, the amount of face-to-face contact in SMS equaled that of the active treatment, and home monitoring of symptoms paralleled the monitoring of relaxation and other skills practice required of the BF/CBT group. Daily monitoring of symptoms and the opportunity to reflect upon and discuss SLE history appeared to have some therapeutic effects, even though no specific information or skills were provided. The effect sizes for pain and disease activity in the SMS condition indicate that some improvements in these areas were observed over time; however, these improvements were not determined to be statistically significant, nor were they of a magnitude that suggests clinical importance. Future studies require larger sample sizes to detect small to moderate changes that may be associated with symptom monitoring and attention.

Although BF/CBT has not been previously studied in SLE, similar programs have been applied to persons with other rheumatic conditions. Our results are similar to those of Buckelew and colleagues (9), who found that a 6-session biofeedback/relaxation program for fibromyalgia led to short-term improvement in pain and self efficacy. Long-term (2-year followup) benefits were maintained only for psychological function variables of depression and self efficacy for managing pain (9). Our findings are also generally consistent with those of Bradley and colleagues (8), who found that a 15-session BF/CBT intervention for rheumatoid arthritis was superior to a support intervention and usual care in the areas of pain and rheumatoid activity index at posttreatment only. Their BF/CBT group maintained benefits at the 6-month followup relative to usual care in the area of trait anxiety. Despite the apparent attenuation of treatment benefits over time in the Bradley study, health care utilization was lower for BF/CBT than other groups 6 and 12 months posttreatment (41). Thus, our finding of limited maintenance of relative improvements over longer-term followup is consistent with previous reports of BF/CBT in patients with other rheumatic conditions, and evaluation of health care utilization in our sample may be warranted.

Previous studies of nonpharmacologic interventions for SLE have included psychoeducational groups (10, 12) or telephone counseling interventions (11, 40) rather than BF/CBT delivered one on one in individual sessions. Comparison of results across studies is complicated by differences in sample selection criteria, structure and content of interventions, and outcome measures. However, the pretreatment–posttreatment effect sizes obtained in the present study for BF/CBT, which ranged from 0.36 to 1.05, compare favorably to those found for telephone counseling interventions, which ranged from 0.14 to 0.54 (11). In addition, the current findings suggest that BF/CBT may be more beneficial than a supportive-expressive group treatment, which was not associated with greater improvement in quality of life or psychological distress than usual medical care alone (12) immediately following treatment. The apparent success of the BF/CBT intervention relative to telephone or group supportive counseling may be due to greater time and attention, or the specific focus on areas of great concern to many persons with lupus, namely, pain and stress reduction.

Differential short-term improvement in the domains of pain, psychological function, and perceived physical function was demonstrated for the BF/CBT condition relative to the control groups. A secondary aim of this study was to explore possible changes in disease activity associated with these interventions. There was not differential improvement in SLE disease activity in the BF/CBT group. Disease activity improved statistically over time in the entire sample, but the groups did not differ significantly in magnitude of change. There are several possible explanations for the disease activity findings. Given the fluctuating nature of SLE, measuring the disease activity effects of an intervention at specific points in time is problematic. In the current study, participants did not receive a medical evaluation of disease activity immediately following the intervention period, rather, these evaluations were conducted 3 and 9 months later. It is possible that differential improvement in disease activity in the BF/CBT group would have been exhibited had medical evaluation immediately posttreatment been feasible. In general, the sample presented with relatively inactive disease initially, which was most likely due to the study's inclusion criteria of stable medication, low to moderate doses of corticosteroids, and no acute renal flare. These criteria were necessary to maximize participation in all sessions and to ensure that improvements could be attributed to the psychosocial treatment rather than medications. However, the resultant sample of individuals with fairly stable disease limits the clinical relevance of any changes in disease activity during the course of the study.

Further investigations of BF/CBT and related behavioral medicine treatment programs for SLE are needed to replicate and expand the results of this study. Future studies should include larger sample sizes to more clearly define differential benefits of comparison interventions. As mentioned earlier, potential treatment-related changes in SLE disease activity over time may have been detected had posttreatment medical evaluations been possible and had we designed the study to include persons with more active SLE disease. The benefits of BF/CBT on pain and function were evident in the short term, but by the 9-month followup evaluation, only psychological function (specifically, self efficacy and perceived stress) remained improved relative to the usual care control group. Future studies should include measures of the impact of lupus on life activities and relationships, such as the illness intrusiveness scale utilized by Edworthy et al (13). Behavioral changes, such as regular practice of relaxation exercises, are difficult to sustain long term without support. Although adherence based on diaries was excellent during the intervention, by the 3- and 9-month assessments less than half of the BF/CBT group reported practicing relaxation “very often.” Future investigations should include postintervention booster sessions to support maintenance of behavioral change. Such support may contribute to longer-term benefits than were observed in the current study.

The study was designed to test a behavioral pain intervention on persons with SLE, which by necessity restricts the sample to persons with pain. Because only SLE patients who reported pain were included, we cannot generalize the results to all persons with lupus. However, pain is present in up to 90% of SLE patients (23, 42). Offering an intervention designed to reduce pain without additional medications may have increased the desirability of the study for some participants. The intervention, although brief, required several clinic visits with a trained health professional. The low attrition rate among those who started the intervention suggests that the program was not overly burdensome. However, for persons with greater limitations on activity or time constraints, minimal contact interventions, such as telephone counseling, may be more feasible.

This study represents the first randomized controlled trial of biofeedback-assisted cognitive-behavioral stress management in persons with SLE. Previous research supports the utility of BF/CBT for improving quality of life and health status in people with other rheumatic diseases, such as rheumatoid arthritis (8) and fibromyalgia (9). In this first test of BF/CBT for SLE patients with pain, the BF/CBT program was predicted to be superior to a symptom monitoring/support condition and usual medical care alone for improving pain, physical function, and psychological function. The results suggest that the program may be a promising adjunct to medical treatment, particularly with regards to short-term effects on pain, psychological function, and physical function and sustained improvements in psychological functioning.