This article requires a subscription to view the full text. If you have a subscription you may use the login form below to view the article. Access to this article can also be purchased.
Objective. We performed the first systematic review with metaanalysis of the efficacy of cognitive-behavioral therapies (CBT) in fibromyalgia syndrome (FM).
Methods. We screened Cochrane Library, Medline, PsychINFO, and Scopus (through June 2009) and the reference sections of original studies and systematic reviews for CBT in FM. Randomized controlled trials (RCT) comparing CBT to controls were analyzed. Primary outcomes were pain, sleep, fatigue, and health-related quality of life (HRQOL). Secondary outcomes were depressed mood, self-efficacy pain, and healthcare-seeking behavior. Effects were summarized using standardized mean differences (SMD).
Results. A total of 14 out of 27 RCT with 910 subjects with a median treatment time of 27 hours (range 6–75) over a median of 9 weeks (range 5–15) were included. CBT reduced depressed mood (SMD −0.24, 95% CI −0.40, −0.08; p = 0.004) at posttreatment. Sensitivity analyses showed that the positive effect on depressed mood could not be distinguished from some risks of bias. There was no significant effect on pain, fatigue, sleep, and HRQOL at posttreatment and at followup. There was a significant effect on self-efficacy pain posttreatment (SMD 0.85, 95% CI 0.25, 1.46; p = 0.006) and at followup (SMD 0.90, 95% CI 0.14, 1.66; p = 0.02). Operant behavioral therapy significantly reduced the number of physician visits at followup (SMD −1.57, 95% CI −2.00, −1.14; p < 0.001).
Conclusion. CBT can be considered to improve coping with pain and to reduce depressed mood and healthcare-seeking behavior in FM.
- FIBROMYALGIA SYNDROME
- COGNITIVE THERAPY
- COGNITIVE BEHAVIORAL THERAPY
- OPERANT BEHAVIORAL THERAPY
- SYSTEMATIC REVIEW
The key symptoms of fibromyalgia syndrome (FM) are chronic widespread pain, fatigue, nonrestorative sleep, and psychological distress1,2. The estimated prevalence of FM in Western European countries ranges from 2.2% to 6.6%3. Comorbidities with affective or anxiety disorders are common4.
Patients with FM use a lot of pharmacological and nonpharmacological therapies, resulting in high costs for health services5. Pharmacological and physical therapies are used more frequently than psychotherapeutic treatments. In an Internet survey, only 8% of respondents reported use of cognitive-behavioral therapies (CBT), but over 80% named emotional distress as an aggravating factor6.
CBT include interventions that are based on the basic premise that chronic pain is maintained by cognitive and behavioral factors, and that psychological treatment leads to changes in these factors through cognitive (e.g., cognitive restructuring) and/or behavioral (e.g., relaxation training, social skills training) techniques. Different types of CBT can be differentiated by the techniques applied. Mindfulness-based stress reduction (MBSR) is a cognitive therapy that helps individuals to self-manage and reframe worrisome and intrusive thoughts by mindfulness meditation, that is, the nonjudgmental awareness of one’s present experience (thoughts, emotions, bodily sensations)7. Operant behavioral treatment (OBT) focuses on the modification of pain behavior by increasing activity levels, and on reducing healthcare-seeking behavior, as well as on including significant others to reduce reinforcement of pain behaviors8. The main therapeutic techniques employed in CBT are the modification of dys-functional thoughts and behavioral modification9.
Systematic reviews and evidence-based guidelines intend to provide a guide through the variety of pharmacological and nonpharmacological treatment options for healthcare professionals and patients. The 3 existing evidence-based guidelines available for management of FM gave different grades of recommendation for CBT. Whereas the American Pain Society10 and the Association of the Scientific Medical Societies in Germany11 gave the highest grade of recommendation to CBT based on qualitative systematic reviews, the European League Against Rheumatism gave only a weak (expert opinion) recommendation for CBT12. The conclusions of narrative systematic reviews on CBT in FM were inconclusive as well. Koulil, et al13 concluded from 6 randomized controlled trials (RCT) that the effects on pain, disability, and mood were limited and that mostly CBT within a multicomponent approach yielded improvements. Bennett and Nelson concluded from 6 RCT that CBT as a single treatment modality did not offer any distinct advantage over well planned group programs of education or exercise, or both9. Thieme and coworkers concluded from 14 studies that CBT were superior to controls in most key domains of FM at posttreatment and followup14. A metaanalysis of the results of RCT with CBT in FM had not been conducted until now. A recent Cochrane Review on the efficacy of CBT in chronic pain syndromes included only 5 studies with FM patients15.
Because of these inconsistent results we saw the need to reexamine the literature and to perform the first quantitative analysis of the outcomes of CBT in FM. The aims of this systematic review were to assess if CBT have beneficial effects at posttreatment and at followup on symptoms of FM compared with controls.
MATERIALS AND METHODS
Methods of analysis and inclusion criteria were specified in advance. We used the review protocol of our systematic review on balneotherapy in FM18.
Types of studies
This was a RCT design comparing CBT with no treatment, treatment as usual, attention control (unspecific elements of CBT such as education, group discussion, or emotional support), or active therapy (any defined pharmacological or nonpharmacological therapy other than CBT). Studies without randomization were excluded.
Types of participants
Patients diagnosed with FM on recognized criteria, of any age, were included.
Types of interventions
RCT with face to face cognitive, operant behavioral, or cognitive-behavioral therapies with defined psychotherapeutic content as an active treatment of primary interest were included. Studies with education only (information on the symptoms and management of FM and/or discussion and/or emotional support) or relaxation and/or biofeedback only were excluded. Studies in which CBT were part of multicomponent therapy were excluded because it would not be possible to separate the effects of CBT from aerobic exercise.
Types of outcomes measures
Studies should assess at least one key domain of FM [pain, sleep, fatigue, patient global multidimensional function, i.e., health related quality of life (HRQOL)]1. Depressed mood, self-efficacy pain (SE Pain, i.e., subjects’ perceived ability to manage and cope with pain and its emotional and behavioral consequences), and healthcare-seeking behavior were chosen for secondary outcomes because emotional status, increasing activity, and coping with pain are the main targets of all types of CBT8,9. Reducing healthcare-seeking behavior is a major goal of operant behavioral therapy8. From each trial we selected the measure considered most appropriate for each of the 7 outcomes. When there was more than one measure for an outcome we gave preference to measures recommended by OMERACT1.
Data sources and searches
The electronic bibliographic databases screened included the Cochrane Central Register of Controlled Trials (CENTRAL), Scopus, Medline, and PsycINFO (through June 30, 2009). The search strategy for Medline is detailed in Table 1. The search strategy was adapted for each database if necessary. No language restrictions were applied. Only fully published reports were reviewed. In addition, reference sections of original studies, systematic reviews9,13,14, and evidence-based guidelines on the management of FM10,11,12 were screened manually.
Two authors (KB, NF) independently screened the titles and abstracts of potentially eligible studies identified by the search strategy as above. The full-text articles were then examined independently by 2 authors (VK, WH) to determine if they met the inclusion criteria.
Data collection process
Two authors (KB, NF) independently extracted the data using standard extraction forms18. Discrepancies were rechecked and consensus achieved by discussion. If needed, a third author (WH) reviewed the data to reach a consensus.
Where means or standard deviations were missing, attempts were made to obtain these data by contacting 5 trial authors. Additional data were provided by 2 authors (Table 2). Where standard deviations were not available from trial authors, they were calculated from t values, confidence intervals, or standard errors, as reported in articles17. If these data were not available, the standard deviation was substituted by the mean of the standard deviations of studies that used the same outcome scale17. In case of different directions of scales the mean from one set of the studies was multiplied by −117.
Data for study settings, participants, exclusion criteria, interventions, cotherapies, attendance rates, reported side effects, and outcomes sought are listed in Table 2.
If studies had 2 or more potential control groups we used the following order to select for control group: attention placebo, treatment as usual, and active control.
Risk of bias in individual studies
To ascertain the internal validity of the eligible RCT, 2 pairs of reviewers (KB, NF; VK, WH) working independently determined the adequacy of randomization, concealment of allocation, blinding of outcome assessors, and adequacy of data analysis (was intention-to-treat analysis performed?). The quality of the treatment was assessed by the 5 items (treatment content/setting, treatment duration, manualization of the treatment, adherence of the therapist to the manual, therapist training and client engagement) of a quality rating scale designed specifically for application to psychological treatment studies in pain. The maximum score is 919. Based on the classification of Yates and coworkers19 we defined scores 0–2 to indicate poor, scores 3–5 average, and scores 6–9 excellent treatment quality. The same pairs of reviewers checked the settings of the studies, the means of referral to the RCT, the inclusion and exclusion criteria, and the socio-demographic data of the study samples and if patients with affective or anxiety disorders were included to assess the representativeness of study samples for FM patients of clinical practice (external validity).
Metaanalyses were conducted using RevMan Analysis software (RevMan 5.0.17) of the Cochrane Collaboration20. Standardized mean differences (SMD) were calculated by means and SD or change scores for each intervention. Examination of the combined results was performed by a random-effects model (inverse variance method), because this model is more conservative than the fixed-effects model and incorporates both within-study and between-study variance21. The SMD used in Cochrane reviews is the effect size known as Hedges (adjusted) g. We used Cohen’s categories to evaluate the magnitude of the effect size, calculated by SMD, with g > 0.2–0.5 = small effect size, g > 0.5–0.8 = medium effect size, and g > 0.8 = large effect size22.
Planned methods of analysis
Heterogeneity was tested using the I2 statistic, with I2 values over 50% indicating strong heterogeneity. Tau-squared was used to determine how much heterogeneity was explained by subgroup differences17.
Risk of bias across studies
Potential publication bias (i.e., the association of publication probability with the statistical significance of study results) was investigated by visual assessment of the funnel plot (plots of effect estimates against its standard error) calculated by RevMan Analysis software if appropriate (at least 10 studies available). Publication bias may lead to asymmetrical funnel plots17.
If there were at least 2 studies available, subgroup analyses were prespecified for type (cognitive, cognitive-behavioral, operant behavioral), duration of total CBT (< 10 h, 10–20 h, ≥ 20 h), type of control group (attention control, no therapy or treatment as usual, active therapy). These subgroup analyses were also used to examine potential sources of clinical heterogeneity.
When at least 2 studies were available, the following sensitivity analyses were prespecified: (1) studies with inadequate or unclear versus studies with adequate sequence generation; (2) studies with inadequate or unclear versus studies with adequate concealment of allocation; (3) studies with no or unclear versus studies with adequate blinding of the outcome assessor; (3) studies without versus studies with intention-to-treat analysis; (4) studies that excluded (or did not report exclusion criteria) versus studies that included patients with affective or anxiety disorders; (5) studies with poor versus average and versus high treatment quality; and (6) because we assumed a growing acceptance of active therapies such as CBT in FM by patients over the years, an analysis with publication of studies before 2000 and then 2000–2005 and 2006–2009 was conducted. These sensitivity analyses were also used to examine potential sources of methodological heterogeneity.
The literature search produced 298 citations; 167 were “double hits” (study found in at least 2 data sources). By screening, 104 records were excluded: 37 did not evaluate CBT in FM, 52 were review articles, and 15 were case reports, conference papers or commentaries. Twenty-seven full-text articles assessed for eligibility; 13 full-text articles were excluded for the following reasons: one lacking a control group23, 2 lacking randomization24,25, one because CBT was only Internet-based26, 5 because CBT were combined with aerobic exercise27,28,29,30,31, one because the data presented in the report were not suited for the predefined outcomes of this metaanalysis and necessary data for meta-analysis were not provided on request32, and 3 because of secondary (e.g., economic) analysis33,34,35 of 2 RCT included in metaanalysis36,37. Finally, 14 studies with 15 study arms were included36,37,38,39,40,41,42,43,44,45,46,47,48,49 (Figure 1).
Setting, referral, and exclusion criteria
Five studies were conducted in North America and 9 in North and Middle Europe. Patients were recruited by registers of hospitals, referral (general practitioner, rheumatologist, departments of hospitals), local self-help groups, and newspaper advertisements. Twelve studies were conducted within the setting of a university, 2 within district hospitals. All studies were single-center based. One study did not report the inclusion and exclusion criteria. Thirteen studies excluded patients with somatic diseases. Seven studies excluded patients with mental disorders. The exclusion criteria of mental disorders were clearly defined in only 5 studies. Three studies excluded patients with unresolved litigation. FM was diagnosed in 14 studies by the criteria of the American College of Rheumatology50 and in one study with adolescents by the Juvenile Primary Fibromyalgia criteria51. Three studies reported somatic comorbidities of the patients. No study performed a psychiatric interview or reported the frequency of mental disorders.
The median of the mean age of participants was 47 years (range 16–54 yrs). The median percentage of women was 100% (range 88%-100%). The median percentage of Caucasians was 91% (range 86%–94%).
Ten studies reported the number of persons screened and randomized with a median of 69% (range 32%–95%). The median number of patients with CBT was 40 (range 7–64) and controls 40 (range 7–63). We found 426/527 (81%) patients in the CBT groups and 288/383 (75%) in the control groups completed therapy (z = −0.9, p = 0.4).
Three studies offered cognitive (MBSR), 2 studies operant behavioral, and 10 studies cognitive-behavioral therapy. All studies but one were outpatient-based. In all studies, CBT were performed as group therapy. Four studies reported the attendance rates with a median of 75% (range 69%–90%).
In 5 studies the controls received treatment as usual or no therapy and in 4 studies they were treated by another active therapy (self-monitoring, aerobic exercise, antidepressants and passive physical therapy, education and low-intensity physical activity). In 6 studies an attention control was used (education, relaxation, and group support and discussion; Table 2).
The length of the interventions, excluding followup, ranged from 5 to 15 weeks (median 9 weeks). The median total treatment time was 27 hours (range 6–75 hours). Twelve studies performed followups. The median of the latest followup was 6 months (range 2–48 months).
The underlying rationale for applying CBT to predefined outcomes was reported by 10 studies. The 3 studies with MBSR aimed to reduce stress and to train mindfulness to reduce the negative effects of thoughts and sensation associated with pain. Primary outcomes were coping with pain and depression. The 2 studies with operant behavioral therapy were intended to increase the physical activity levels at home and at work, to reduce healthcare-seeking behavior, and to include significant others to reduce reinforcement of pain behaviors. The primary outcomes were disability, pain, affective distress, and number of physician visits. The rationale reported by 4 studies with CBT was to teach patients the skills needed to control pain and disability and to enhance self-efficacy pain. The primary outcomes used most frequently by these studies were pain, disability, and self-efficacy pain. One CBT study focused on improvement of sleep to interrupt the pain/distress cycle. Outcomes of this study were pain, sleep quality, depressed mood, and HRQOL.
There was a great variety of most outcomes measures. Some studies did not assess all outcomes of the review. The subscales of the Fibromyalgia Impact Questionnaire were not reported by 2 studies and were not provided on request (Table 2). No study assessed predefined response rates (e.g., percentage of patients with 30% pain reduction). Other potential outcomes of interest, e.g., anxiety or physical function, were used by only a minority of studies.
Four studies reported on side effects. Three studies reported dropouts because of increase of symptoms in the CBT group (5/129 patients) and one study in the control group (2/39 patients). One study reported that no treatment-related side effects occurred.
Risk of bias within studies
One study fulfilled all predefined criteria for methodological quality (Table 3). The reported quality of treatment was low in most studies (Table 4). Seven studies included patients with mental disorders.
Results of individual studies
The means, SD, sample sizes, and effect estimates at posttreatment of each study can be seen in the forest plots (Figure 2).
Synthesis of results
Data are reported as follows: standardized mean difference, 95% confidence interval; p value of test for overall effect.
CBT reduced depressed mood (0.24, 95% CI −0.40, −0.08; p = 0.004) and improved self-efficacy pain (0.85, 95% CI 0.25, 1.46; p = 0.006) compared to controls at posttreatment. There was no significant effect on fatigue (−0.09, 95% CI −0.27, 0.51; p = 0.61), sleep (−0.15, 95% CI −0.60, 0.29; p = 0.50), and HRQOL (−0.13, 95% CI −0.40, 0.15; p = 0.37). Based on Cohen’s categories the effects were small for depression and high for self-efficacy pain.
CBT improved self-efficacy pain (0.90, 95% CI 0.14, 1.66; p = 0.02), and operant behavioral therapy reduced the number of physician visits (−1.57, 95% CI −2.00, −1.14; p < 0.001) compared to controls at the latest followup. These effects were high. There were no significant effects on pain (−0.20, 95% CI −0.57, 0.16; p = 0.28), fatigue (−0.33, 95% CI −0.87, 0.21; p = 0.23), sleep (−0.30, 95% CI −1.04, 0.44; p = 0.44), and depressed mood (−0.16, 95% CI −0.35, 0.02; p = 0.07) at latest followup (Table 5).
Risk of bias across studies
There was substantial heterogeneity in all comparisons of outcomes at posttreatment and at latest followup, except for fatigue and depressed mood at posttreatment and for depressed mood at followup (Table 5).
On visual inspection, the funnel plots of the outcomes pain, depressed mood, and HRQOL did not indicate publication bias (data available on request).
None of the comparisons of subgroup analyses yielded a significant test for overall effect for the outcome of pain. Statistical heterogeneity of analysis was substantially reduced in case of MBSR and CBT, in case of duration < 10 and 10–20 hours, and if attention placebo and therapy as usual/no therapy were used for controls. Heterogeneity was due to cases of operant therapy, therapies > 20 hours, and active therapies as controls (Table 6).
Stratification according to potential risks of methodological bias for the outcome with a significant test for overall effect, namely depressed mood at posttreatment, confirmed only partially the robustness of the results: the test for overall effect was significant only in studies without adequate allocation concealment, without intention-to-treat analysis, with moderate treatment quality, without inclusion of patients with affective and anxiety disorders, and with adequate randomization. Only studies published in the period 2006–2009 had a significant test for overall effect on depressed mood. Statistical heterogeneity of analysis was substantially reduced in all comparisons (Table 7).
Summary of evidence
There was evidence of the efficacy of CBT to reduce depressed mood at posttreatment and to improve self-efficacy pain at posttreatment and at followup. The positive effect on depressed mood cannot be clearly distinguished from biases. There was evidence of the efficacy of operant behavioral therapy to reduce the number of physician visits at followup. There was no evidence of the efficacy of CBT to reduce pain, fatigue, sleep disturbances, and limitations of HRQOL at posttreatment and at followup.
Applicability of evidence
The study settings of secondary and tertiary care centers and the study samples, with a preponderance of middle-aged women, are representative for populations with clinical FM in the USA and Northern and Middle Europe.
Although every effort was made to obtain missing data from authors, it was not possible in every case to do so. Therefore some studies are not represented fully in our metaanalysis. We decided to substitute missing data by means of other studies using the same outcome scale despite the small sample sizes and substantial heterogeneity in some outcomes because we intended to include all available studies into this metaanalysis.
The methodological quality of the studies varied. Considerable heterogeneity existed for the outcomes of pain, sleep, fatigue, HRQOL, and self-efficacy pain posttreatment and at followup, and this could mainly be explained by clinical and methodological differences between the studies.
Formal blinding of participants and clinicians to treatment arm is not possible in psychotherapy trials. Therefore underestimation of the extent to which clinicians’ and participants’ knowledge of group assignation influenced the true effect could be possible.
The reliability of the results of some sensitivity and subgroup analyses is limited because the analyses were under-powered due to the small number of included studies.
Responses in studies in patients with chronic pain are frequently not Gaussian, but with a split between responders and nonresponders. No study assessed predefined response rates (e.g., 30% pain reduction). Therefore the IMMPACT response outcomes (Initiative on Methods, Measurement, and Pain Assessment in Clinical Trials52) could not be calculated.
There is no “gold standard” for the methods of a meta-analysis, e.g., for combining results and assessing risks of bias within and between studies. We followed a recent recommendation on the methods of a systematic review with metaanalysis16. We used the same outcomes and statistical methods as a recent Cochrane Review on psychological therapies in chronic pain15 and other systematic reviews of the German guideline group on pharmacological and nonpharmacological therapies in FM18,53,54,55 to allow indirect comparisons of the results of systematic reviews.
Agreements with other systematic reviews
Our metaanalysis does not confirm the conclusion of a qualitative systematic review of the German FM guideline group that CBT are superior to controls in most key domains of FM at the end of therapy and at followup14. Our results are partially in agreement with a recent Cochrane Review on CBT and behavioral therapies in chronic pain syndromes, which found that both were effective in altering mood outcomes at the end of treatment. In contrast, we previously could not find a weak effect on pain in patients with FM15. Therefore the high grade of recommendation given to CBT in the American and German guidelines on FM10,11 needs to be reconsidered.
Agreement with excluded studies
The lack of efficacy of CBT on most key symptoms of FM is mainly confirmed by the studies we excluded. One RCT found no differences between CBT and treatment as usual on pain32. One RCT with the Internet-based arthritis self-management program found it was not superior to usual care on pain, fatigue, and disability at the 1-year followup26. One controlled study found CBT was superior to waiting list on pain, fatigue, and HRQOL, but not on sleep25. Another controlled study found no differences between CBT and nontreated patients on sleep and depressed mood at posttreatment24. One CBT study37 reported the effects on healthcare use at followup in a second publication32; CBT was not superior to education alone. We excluded this study from our metaanalysis because no specific interventions for healthcare use were delivered and pain behavior was not expected to be affected by the program37.
Agreement with other studies on psychological therapies in FM
An increase in self-efficacy pain predicted the reduction of pain intensity and disability at followup by multicomponent therapy including CBT56. In contrast, the large effect size of CBT on self-efficacy pain in this metaanalysis was in contrast to the lack of effects of CBT on pain and HRQOL measures (which included items of disability).
Implications for clinical practice
CBT cannot be recommended for therapy of the key symptoms of FM, namely pain, fatigue, sleep disturbances, and HRQOL. The lower rate of side effects and dropouts in the studies of CBT compared to antidepressants53 can be a major argument for CBT as a treatment option for depressive symptoms. Operant behavioral therapy can be considered to reduce healthcare-seeking behavior of patients with FM.
Implications for research
The methodological quality of further studies could be improved by consideration of the following issues. The use of a core set of outcome measures including response rates would improve the internal validity of a metaanalysis52. Recommendations on the quality of the treatment delivered, study design, and statistical methods should be followed19. Potential side effects, such as worsening of symptoms or increase of interpersonal conflicts, should be documented to allow indirect comparisons of the safety of CBT versus pharmacological therapies. More studies with adolescents and separate studies or subgroup analyses of male patients are needed to clarify the effects of CBT in these patients. Predictors of positive treatment outcomes, e.g., improved self-efficacy pain, should be investigated. Further studies are required to determine if CBT tailored to subgroups of patients with FM (e.g., with and without affective disorder) will improve the key symptoms of FM.
- Accepted for publication May 26, 2010.