2.1 Study population

The study was performed in a municipality in central Sweden with approximately 22,000 inhabitants in 2001–2003. The study population was recruited by advertising in the local daily newspaper and an information meeting with the local branch of the Fibromyalgia Patient Association.

Responding female women with FMS were invited to an examination at the coordinating primary health care centre. Inclusion criteria were age 18–64 years, being Swedish-speaking, and fulfilment of the 1990 ACR criteria [1] (generalized pain for more than three months, distributed in all four body quadrants, and at least 11 tender points in typical locations) . Exclusion criteria were major psychiatric or somatic disease, and substance abuse.

Information was sought on duration of generalized pain as well as time since possible FMS diagnosis. Tender points were assessed manually by finger top pressure of 40N/cm² by one physician (BK) . History of severe psychiatric or somatic disease was obtained from medical records, in addition to information from the patient during the screening examination. No formal testing of psychological disorders or symptoms was done during this examination. A physiotherapist experienced in FMS tender point assessment validated the diagnostic procedure.

Among the 54 female patients recruited, six patients were excluded after these procedures. Two did not fulfil the diagnostic criteria, two had a serious mental disorder, and two declined participation after receiving further information about the study. The remaining 48 women agreed to participate and were using a random block design allocated into two groups, group 1 (n = 24) and group 2 (n = 24) . The randomization was performed with the SAS function ‘ranuni’ that produces random numbers with equal distribution, i.e. all numbers appear with the same probability. According to this design for every four consecutive patients two were randomly allocated to group 1 and the remaining two were allocated to group 2. The allocations were indicated on paper sheets and put in sealed envelopes with a patient serial number on the outside. The sheet furthermore had a disturbing text on the backside to prevent reading the allocation through the envelope. The envelopes were stored with the study monitor. When patients were included in the study they were given a serial number, the corresponding serial number envelope was opened and the patient allocation was noted in the study chart.

2.2 Study design

The study was designed as a two-arm parallel group RCT, with group 1 as intervention group and group 2 as control group (Fig. 1) . The intervention group received the CBT treatment and the control group was wait-listed. When the RCT was concluded, group 2, the previous control group, received the same CBT treatment as the group 1. A before-and-after treatment design was then performed for the two groups together (increasing the number of treated subjects from 24 to 48) , a standard procedure in RCTs with small study populations.

Fig. 1

Flow chart of the study population.

The patients’ local physicians were informed about the study and were responsible for the every-day care of the patients. There were no restrictions in changing medication or other treatment modalities. However all such changes were assessed and documented in the study protocol. All participants gave oral informed consent to participation, standard procedure at the time, and the trial was performed in accordance with the Helsinki declaration. The Research Ethics Committee at Uppsala University approved the study.

2.3 Data collection

Data was collected on three occasions for group 1 and four occasions for group 2 (Fig. 1) . The first measurement was performed after randomization but before the treatment or wait list period began. Measurements were then made after the 6 and 12 months in group 1 and after 6, 12 and 18 months in group 2, For examination and data collection the patients attended the study coordinating primary health care centre between 8:00 and 10:00 a.m. For premenopausal women all measurements were performed 9-14 days after the first day of the last menstrual period to avoid interference regarding endocrine variables [19]. In order to obtain reliable self-reported data, it was emphasized that the study would have no consequences for disability or sick-leave compensation.

2.4 Intervention

The CBT stress management programme is described in detail in the electronic supplementary material. Briefly, the intervention consisted of twenty CBT sessions during a 6-month period. The women were allocated into treatment subgroups with 5–7 women in each, receiving one group session a week, with 3-hour duration. When the 20 sessions were completed, three CBT group booster sessions were given during the next six months, each with three hours’ duration. Two psychologists trained in CBT provided the treatment. To secure continuity, each subgroup had the same therapist throughout the programme. The psychologists took no part in planning or evaluation to avoid interference with the results of the study. Median attendance rate was 93%; inter quartile range 91–100%.

Quality assurance of the delivery of treatment was achieved through continuous supervision by one of the co-authors (GB) . Treatment sessions were also monitored by patients’ subjective evaluations of pain, stress and well-being with visual analogue scales (1–100 mm) , which gave an indication of the feasibility and attractiveness of the treatment from the patients’ perspective.

The overall goal of the CBT was to develop emotional, behavioural, and cognitive coping strategies for dealing with stress and pain. Treatment components included knowledge, self-monitoring, behavioural skills training, cognitive restructuring, and life value issues. Life values are important determinants of motivation for behaviour change, and for the individual’s own purpose and long-term goals and maintenance. Dealing with life values can be subsumed under ‘cognitive restructuring’.

The therapeutic material included case illustrations, audio-visual material, readings, hand-outs, exercises, and thematic discussions. Items of special importance, such as anxiety, exhaustion, self-esteem, and marital distress were discussed [33]. Homework assignments were applied between each session and included self-monitoring by simple diaries as well as a booklet with behavioural and cognitive exercises.

Each session had an agenda and a specific theme. The current theme was discussed and elaborated, and new themes and issues were introduced, building on previous discussions. A short relaxation technique (Jacobsen’s progressive relaxation technique) was taught [39].

2.5 Statistical analysis

The statistical analyses were conducted using the SAS software, version 9.3 [40]. Partial non-response (missing data in returned questionnaires) was on average 0.6% with a maximum in individual variables of 1.1%. Simple differences between groups in continuous variables were tested with Student’s t-test, and differences in proportions with the chi-square test.

Two analytical designs were used, both based on the intention-to-treat approach. The first, a classical RCT, was based on a traditional two-arm parallel group design with group 1 serving as intervention group and group 2 as control group. The second design was based on a before-and-after treatment approach in both groups after intervention. The purpose with the two designs was to use data more efficiently and to crosscheck results from the two designs. Primary outcome was the MPI-1 dimension ‘life control’, secondary outcomes were the MPI-1 dimensions ‘interference’, ‘affective distress’ and ‘support from spouses or significant others’, the various MPI-2 dimensions, the ‘general activity level’ in MPI-3 dimension, and ‘vital exhaustion’, ‘stress behaviour’, and ‘depression’. The only tertiary outcome was ‘pain severity’ in the MPI-1.

No a priori power analysis was made, but a post hoc analysis was done, based on the RCT design covering the first 6 months. For the primary outcome ‘life control’ the power to detect the differences found between group 1 and 2 was 90% given the size of the study population, contrasting to the power of 38% found for the one of the secondary outcome ‘stress behaviour’.

The RCT analyses were performed with the SAS procedure ‘General Linear Model’ (linear regression) and its option Repeated Measures Analysis of Variance. The baseline and final levels of the outcome variable were used as the dependent variables, and the group assignment as the independent variable. The repeated measures analysis does not assume uncorrelated measures across time. The before-and-after analyses were performed accordingly, with groups 1 and 2 pooled and group assignment used as covariate (variance reducer) .

The analyses were performed in two steps. First, preliminary analyses were performed with outcome as the dependent variable and the group variable and each of the possible determinants, one at a time, as independent variables. Significant variables in the preliminary analyses were entered into the final multivariate analyses.

The final analyses of the RCT design were performed as multivariate analyses as described above, but in addition, the initial outcome level (to adjust for bias due to different initial levels between groups) and variables affecting outcome, other than the treatment groups and initial outcome level, identified in the preliminary analyses, were entered as covariates. The latter were ‘being in gainful work’, ‘waist/hip ratio’, ‘smoking habits’, ‘alcohol consumption’, ‘duration of generalized pain’, ‘number of tender points’, ‘menstrual status’, ‘important childhood and adolescence experiences’, and ‘therapist’. To avoid model overload, backward elimination of non-significant variables was used. Also stepwise entrance of independent variables was used but gave results almost identical to backward elimination. For this reason the latter was used.

The final analysis in the before-and-after treatment design was simpler, since both groups served as their own controls. The only covariates used were group assignment and initial outcome level. No significant interactions were found between group assignment, outcome variables and time. All tests were two-tailed. The level of significance in the preliminary analyses was set at p < 0.10 and in the final analyses models at p < 0.05.

3.1 Characteristics of the study population

Characteristics of the study population are shown in Table 1. Mean age was 49 years, about 85% were married, 25% had mandatory education only, less than half were in gainful work, mean body mass index was 29, and approximately one out of ten were smokers. The average alcohol consumption was 5 g of pure alcohol during the last week, and the mean physical activity score on a four-point scale was 1.9. The patients had on average a pain history of more than ten years, more than five years with a fibromyalgia diagnosis, and on average 16 out of 18 diagnostic tender points. About 40% were on anti-depressants, and more than half were on intermittent or continuous analgesics, slightly less than half were on hormone replacement therapy. There were no significant differences between the two groups in these variables. Two CBT therapists gave treatment in group 1versus one in group 2.

No significant outcome differences were observed between the assessments of the two treatment psychologists. During the study, 10 participants were started on anti-depressant medication and 6 discontinued.Taking or changing the intake of anti-depressants did not influence the outcome of this study. No non-pharmacological treatment for FMS by us or by others was administered during the study period. There were no differences in number of tender points.

3.2 Outcome in the two-arm group parallel design (RCT)

Outcome variables of the RCT design are presented in Table 2, with crude levels and levels adjusted for potential covariates and baseline differences of the outcome variables in group 1 before- and-after treatment and the corresponding points in time for group 2 before-and-after the wait list period.

In the ‘psychosocial dimension’ scales (MPI-1) , ‘life control’ score increased in group 1 from 3.15 to 3.62, while it decreased in group 2 to 2.86 (p = 0.01) , ‘pain severity’ score increased from3.61 to 4.20 in group 1 and decreased to 3.37 in group 2 (p = 0.02) and ‘interference’ score increased from 3.70 to 4.07 in group 1 and decreased to 3.45 in group 2 (p = 0.04) . ‘Affective distress’ score declined from 2.75 to 2.20 in group 1 as compared to a rise to 3.08 in group 2 (p = 0.01) . The ‘support from spouses or significant others’ score increased from 3.64 to 3.92 in group 1 as compared to a decline to 3.03 in group 2 (p = 0.03) .

In the ‘behaviour – significant others dimension’ (MPI-2) there was a significant increase in group 1 as compared with group 2 for the ‘distracting’ responses score (p = 0.03) . There were no significant differences between groups in the ‘behaviour - activities dimension’ (MPI-3) . ‘Depression’ score decreased in group 1 while it increased in group 2 (p = 0.02) . For the ‘sleep’ score in MADRS-S the differences in trend between groups were non-significant.

3.3 Outcome in the before-and-after treatment design

Outcome variables for the before-and-after treatment design are presented in Fig. 2. ‘Life control’ improved almost 20% during therapy and follow up (p <0.01) , ‘affective distress’ improved by 15% (p < 0.05) , while the ‘interference’ and ‘support from spouses or significant others’ showed no tendency towards change. ‘Vital exhaustion’ improved by 12% (p < 0.01) , ‘stress behaviour’ improved by 15% (p < 0.01) .Total ‘depression’ score improved by 20% (p <0.01) , while differences in the ‘sleep’ component were non-significant. For ‘pain severity’ there was a non-significant trend towards improvement. For life control-; ‘affective distress-; ‘vital exhaustion’, ‘stress behaviour’ and ‘depression’ there were an improvement not only during the six-month trial period, but also during the six-month follow up. There were no effects of analgesic, antidepressant or hormone replacement therapy in any of the designs.

Fig. 2

Combined results according to the before and after treatment in group 1 (n = 24) and before and after treatment in group 2 (n = 24) , altogether 48 patients.

4.1 RCT and the before-and-after design

The results from the RCT and the before-and-after treatment designs were fairly consistent. Positive effects in the RCTfor ‘life control’, ‘affective distress’, and ‘depression’ were also seen in the before-and-after treatment design during the treatment period and tended to be further enhanced at follow up 12months after start of CBT treatment. The subjective impairment in the RCT for ‘pain severity’ and ‘interference’ from pain was not observed in the before-and-after treatment design, nor were perceived ‘support from spouses or significant others’ or ‘distracting’ responses.

‘Vital exhaustion’ and ‘stress behaviour’ ratings were reduced for both groups in the RCT but the difference between groups were non-significant. In the before-and-after treatment design a small but significant improvement, enhanced during the follow up period, was seen, possibly an effect of higher statistical power in the before-and-after design than in the RCT, but this may also reflect that changes of long established behaviour may take some time to be implemented.

In the RCT, ‘pain severity’ and ‘interference’ from pain both increased more in the intervention group than among controls, while ‘life control’ showed the opposite pattern. A possible expla-nation might be that the women reflected more on their pain and the impact of it, in line with the treatment goals of self-monitoring aimed to increase the awareness of ones’ own reactions, and at the same time they gained more control over their lives and pain behaviour in general. Since the CBT programme focused on stress in relation to pain, and included diaries and self-monitoring, women might have increased their awareness of pain sensations. However, the results indicate that the negative impact of pain in daily life decreased.

The findings are consistent with the aim of the CBT, focusing on stress behaviour and general well-being in favour of techniques of direct pain reduction. The improvement towards more adequate pain behaviour strategies and well-being is in line with earlier experiences from CBT treatment, for instance in coronary patients where differences between CBT treated and non-treated patients were maintained and enhanced up to seven years after the conclusion of the treatment period [34].

The changes found in the present study were numerically rather small and may not always have been considered clinically significant in previous studies. However, given that these patients often have suffered from their disorder for years, even small changes may be interpreted as beneficial, and even more so if the positive changes from various variables co-operate and tend to persist for a long time after therapy is concluded.

4.2 Previous studies

The intervention in this study was based on a group based stress management CBT, tailored for treatment of patients with cardiovascular diseases has proven to be very effective in reducing recurrent heart attacks [41].

The implications of CBT as a treatment facility in fibromyalgia is reflected, apart from the present study, in a meta-analysis of psychological treatments for fibromyalgia made by Glombiewski et al. [42]. The inclusion criterion in the meta-analysis was that at least three out of five outcomes (pain, sleep, depression, catastrophizing, and functional status) were used to evaluate the efficacy of the treatment. They identified 23 eligible studies including 30 psychological treatment conditions and 1396 patients. Of these eight used CBT as one treatment modality [43,44,45,46,47,48,49,50]. They drew the conclusion that psychological treatment in fibromyalgia may be effective interventions and the effects comparable to those in short-term drug treatments. The stable long-term effects and larger effect size of CBT and other psychological interventions indicate that these treatments are more favourable than other non-psychological treatments, which only give short-term effects [51]. When patients attribute positive changes to their own effort, the effects are usually maintained better than effects attributed to external causes, such as medication [51,52,53,54].

Acceptance and commitment therapy, one of the recent developments of CBT was studied in a randomized clinical trial by Wicksell et al. [55]. Pain-related functioning and psychological flexibility improved in spite of no reduction in pain intensity. In a Cochrane Collaboration review including 23 studies with 2031 patients the effects of CBT on FMS was evaluated. It was concluded that CBT, as compared to control treatments, might to some extent reduce pain, negative mood and disability at the end of treatment and after follow-up. Pain was rated with a 6.3% absolute improvement. However among the ten studies with an intention-to-treat design only four showed a significant result [56].

4.3 Therapeutic implications

Thus, in chronic pain treatment, assessment of beliefs, appraisals, and coping strategies are prerequisites for optimizing a treatment programme [51,57]. The cognitive-behaviour framework implies a treatment focus on changes of maladaptive behaviour and cognitive coping in the direction of a more adaptive one, and an increase in self-efficacy and sense of control. Research findings indicate that a treatment approach focusing on development of strategies for coping with pain may result in a decrease of the intensity of the pain experience, as well as reduced psychological distress [51].

Given that total abolishment of pain symptoms is extremely difficult or impossible to achieve, interventions should focus on the effects of pain on daily function and quality of life, and thereby give the patients more satisfaction and possibly a reduction of the pain experience. Thus, the development of individual strategies for coping with pain is essential to reduce its impact on daily life. In order to achieve increased self-efficacy and sense of internal control over one’s life, the individual’s own motivation (reason and need) for change should be an important target for intervention. Since stress may worsen the pain experience, coping with stress might be a promising route to accomplishing that goal.

It is well known that psychological and social factors influence perception and perpetuation of chronic pain [58,59]. It has been suggested that patients with FMS are more affected by normal daily stress than healthy individuals [11,60]. FMS patients may be caught in a vicious circle where pain and negative emotions may lead to fears of more pain. Cognitive factors, such as intrusive thoughts (‘catastrophizing’) , may act as discriminative stimuli for avoidance behaviours [58]. The risk of developing behaviour patterns that gradually decrease activity and daily function is particularly great when there is no known cause of the pain symptom [57,58,59]. Intrusive thoughts are powerful triggers for avoidance behaviour patterns, and may be reinforced by increased attention from significant others and health care staff [61]. It has been shown that sense of self-efficacy, i.e., belief in one’s ability to act and influence outcomes of one’s actions, increases as intrusive thoughts diminish [51,62].

4.4 Strengths and limitations

The strengths of the present study include that only validated instruments were used for outcome analyses, most of which have been used extensively in previous studies. The intervention CBT protocol used in this study has been validated and used in a number of previous studies [34,41,63]. Furthermore, the participation rate was high, no differences in outcome attributable to which of the two CBT therapists that gave the therapy, and the CBT therapists were not involved in the outcome assessments. The monitoring of the study was excellent, resulting in very small data loss. The statistical analysis was state of the art, and the results from the two analytical designs were fairly consistent. Moreover, in the RCT potential outcome affecting variables other than the CBT were taken into account, optimizing analytical precision and conclusion validity. This was not necessary in the before-and-after design, since the subjects were their own controls.

One further strength is that the data for pre-menopausal women were collected 9–14 days after the first day of the last menstrual period to avoid interference regarding endocrine variables [19]. In order to obtain reliable self-reported data, it was emphasized to the participants that the aim of the treatment was not to push them to a higher level of working capacity, but to increase their well-being and quality of life.

The limitations of the study include that the study population was fairly small. However, we had so many participants in the study as was deemed feasible for financial and logistic reasons. No a priori power analysis was made, but a post hoc analysis was done, showing that the study population was large enough for the main hypothesis testing. The study population was not necessarily representative of FMS patients in the general population, even though the vast majority of FMS patients in Sweden are treated in general practice. The patients’ own physicians carried out the usual patient care with no restrictions in terms of changing medication or other treatment modalities. However, all such changes were registered in the study protocol. Another possible limitation might be that no specific measures examining pain appraisal were included, allowing examination of for instance pain catastrophizing and hyper vigilance.