Tue, 3 Apr 2001 The Lancet Vol 357; 9259, pp 841-847 March 17, 2001 URL: http://www.sciencedirect.com/science/journal/01406736 Fig. 1-3: http://www.sciencedirect.com/cache/MiamiImageURL/B6T1B-42M1F22-C-1/0 http://www.sciencedirect.com/cache/MiamiImageURL/B6T1B-42M1F22-C-3/0 http://www.sciencedirect.com/cache/MiamiImageURL/B6T1B-42M1F22-C-5/0 Cognitive behaviour therapy for chronic fatigue syndrome: a multicentre randomised controlled trial ----------------------------------------------------------------------- Judith B. Prins(a), Gijs Bleijenberg(a), Ellen Bazelmans(a), Lammy D. Elving(b), Theo M. de Boo(c), Johan L. Severens(d), Gert Jan van der Wilt(d), Philip Spinhoven(e) and Jos W. M. van der Meer(b) a Department of Medical Psychology, University Medical Centre Nijmegen, PO Box 9101, 6500 HB Nijmegen, Netherlands b Internal Medicine, University Medical Centre, Nijmegen; and Department of Psychology and Psychiatry, Leiden University, Leiden, Netherlands c Medical Statistics, University Medical Centre, Nijmegen; and Department of Psychology and Psychiatry, Leiden University, Leiden, Netherlands d and Medical Technology Assessment, University Medical Centre, Nijmegen; and Department of Psychology and Psychiatry, Leiden University, Leiden, Netherlands e University Medical Centre, Nijmegen; and Department of Psychology and Psychiatry, Leiden University, Leiden, Netherlands Corresponding author. Department of Medical Psychology, University Medical Centre Nijmegen, PO Box 9101, 6500 HB Nijmegen, Netherlands; email: j.prins@cksmps.azn.nl Available online 19 March 2001. Abstract Background/ Cognitive behaviour therapy (CBT) seems a promising treatment for chronic fatigue syndrome (CFS), but the applicability of this treatment outside specialised settings has been questioned. We compared CBT with guided support groups and the natural course in a randomised trial at three centres. Methods/ Of 476 patients diagnosed with CFS, 278 were eligible and willing to take part. 93 were randomly assigned CBT (administered by 13 therapists recently trained in this technique for CFS), 94 were assigned the support-group approach, and 91 the control natural course. Multidimensional assessments were done at baseline, 8 months, and 14 months. The primary outcome variables were fatigue severity (on the checklist individual strength) and functional impairment (on the sickness impact profile) at 8 and 14 months. Data were analysed by intention to treat. Findings/ 241 patients had complete data (83 CBT, 80 support groups, 78 natural course) at 8 months. At 14 months CBT was significantly more effective than both control conditions for fatigue severity (CBT vs support groups 5.8 [2.2-9.4]; CBT vs natural course 5.6 [2.1-9.0]) and for functional impairment (CBT vs support groups 263 [38-488]; CBT vs natural course 222 [3-441]). Support groups were not more effective for CFS patients than the natural course. Among the CBT group, clinically significant improvement was seen in fatigue severity for 20 of 58 (35%), in Karnofsky performance status for 28 of 57 (49%), and self-rated improvement for 29 of 58 (50%). Prognostic factors for outcome after CBT were a higher sense of control predicting more improvement, and a passive activity pattern and focusing on bodily symptoms predicting less improvement. Interpretation/ CBT was more effective than guided support groups and the natural course in a multicentre trial with many therapists. Our study showed a lower proportion of patients with improvement than CBT trials with a few highly skilled therapists. Introduction Chronic fatigue syndrome (CFS) is characterised by persistent or relapsing unexplained fatigue, of new or definite onset and lasting for at least 6 months. Fatigue is not the result of an organic disease or ongoing exertion, rest does not alleviate it, and there is substantial limitation of occupational, educational, social, and personal activities [1]. No cause of CFS has been found, and most patients do not recover. No somatic or pharmacological treatments have proven to be effective [2]. Cognitive behaviour therapy (CBT) seems to be a promising treatment of CFS [3, 4 and 5]. Two randomised controlled trials reported positive results [6 and 7]. A recent review [2] questioned whether these results can be generalised outside specialist centres where only a few highly skilled therapists, or even a single therapist [7], administered CBT. Furthermore, in both studies the primary outcome variable was functional impairment and not fatigue, the main complaint of CFS patients. In our study, criticisms of both previous randomised trials were addressed. The effectiveness of CBT was tested in a multicentre randomised trial. CBT was compared with a treatment condition, guided support groups, and a control condition, the natural course. CBT was administered in three different centres rather than one specialist centre. Experts taught the treatment protocol to many therapists with no previous experience in CBT for CFS. Guided support groups should control for the absence of specific cognitive-behavioural interventions and the presence of therapist's attention and treatment expectations. We assumed that support groups, as in other chronic diseases [8 and 9], might contribute to a feeling of mutual understanding, acceptance, and support, and thereby have a healing effect. In this study, the outcome variables were fatigue severity and functional impairment, with the same instruments used for inclusion and outcome. Moreover, CBT for CFS was based on a statistically tested model of perpetuating factors in CFS[10 and 11] rather than on hypothesised factors in CFS or on treatments of other medically unexplained syndromes. The model of CFS is shown in Fig. 1. Focusing on bodily symptoms, low physical activity, and low sense of control contribute to increasing severity of fatigue and functional impairment. CBT is directed at these perpetuating factors. The main aim of our multicentre trial was to show the effectiveness of CBT for patients with CFS. Our hypothesis was that fatigue severity and functional impairment should decrease significantly more in the group of patients assigned CBT than in patients in the control groups. Patients and methods PATIENTS All patients with a major complaint of fatigue referred to the outpatient clinic of the departments of internal medicine of the University Medical Centre Nijmegen and the University Hospital Maastricht between October, 1996, and December, 1998, were assessed by means of detailed history, physical examination, and computer assessment of questionnaires. Patients were eligible for the study if they met the US Centers for Disease Control and Prevention criteria for CFS [1], with the exception of the criterion requiring four of eight additional symptoms to be present. Severe fatigue and severe functional impairment were defined by cut-off scores - a score of 40 or more on the subscale fatigue severity of the checklist individual strength and a score of 800 or more on the sickness impact profile. Additional inclusion criteria for this study were age between 18 and 60 years and residence within 1.5 h travelling time of one of the study centres. Additional exclusion criteria were previous or current participation in CFS research, pregnancy, and current treatment to achieve pregnancy. A sample size of 80 patients per group was estimated assuming significance of 5%, power of 90%, a dropout rate of 20%, and a medium effect size on the actometer, the measure of our multidimensional approach in need of most individuals to show improvement. Multidimensional assessment has been recommended for studies assessing the effect of therapeutic interventions for CFS, to measure change in different dimensions of the patients' functioning [12]. During the trial, the dropout rate was higher than that estimated in the calculation of sample size. Therefore, the target sample size for inclusion was set at 90 patients per study group. DESIGN AND PROCEDURES The study was an open multicentre randomised controlled trial in which individual CBT was compared with participation in guided support groups and with the natural course, a control condition in which no treatment was offered. The ethics committees of the three participating centres gave approval for the study. Treatment effects were expected in the primary outcome variables fatigue severity and functional impairment and were explored in the secondary outcome variables: Karnofsky performance status, psychological wellbeing, quality of life, and work. The predictive role of perpetuating factors in the model of CFS was tested exploratively also. Patients who met the trial criteria and were willing to take part in the trial had to give informed written consent. To ensure adequate generation and adequate concealment in the allocation process [13], patients were allocated sequentially to one of three conditions, by blockwise randomisation (block size six), separately for each centre. The allocation was concealed in series of envelopes for each centre and assigned by (assistant) researchers before baseline in the presence of the patient, in order of enrolment in the trial. CBT and support groups took place in three different settings, the Department of Medical Psychology of the University Medical Centre Nijmegen, the Department of Psychiatry of the Leiden University Medical Centre, and the Department of Psychotherapy of the Maastricht Mental Health Institute. CBT and support groups were administered by different therapists and on different days to prevent contamination. CBT consisted of 16 sessions of 1 h over 8 months. Patients in this group had to meet the requirements of no further medical examinations or other treatments for CFS during the trial. These conditions were essential in reducing focusing on bodily symptoms and somatic attributions. A preliminary version of CBT has been extensively described [11]. An essential part of CBT is self-control: this means that the CFS patient is acquiring control over symptoms instead of dependence on physicians prescribing treatments or medications. In this study, CBT was outlined in a treatment protocol. First, the model of perpetuating factors was explained, and the therapist attempted to motivate the patient for CBT. Next, fatigue-related cognitions were challenged to diminish somatic attributions, to improve sense of control over symptoms, and to facilitate behaviour change. Patients were encouraged to attain and maintain a base level of physical activity needed to prevent bursts of activity and resultant extreme fatigue. Subsequently, a structured activity programme was started. After a gradual increase of physical activity, a plan for work rehabilitation was outlined and carried out. For patients without a job, rehabilitation in other personal activities was achieved. The final sessions dealt with relapse prevention and further improvement of self-control. 13 behaviour therapists of three different disciplines (psychologists, psychiatrists, and health scientists) took part. Therapists varied in previous CBT (non-CFS-related) experience, because the study was done with the therapists available within the three centres. However, none of the therapists was familiar with CBT for CFS at the start of the trial. Two experts in CBT for CFS (GB, EB) trained the therapists in using the treatment protocol in a workshop, consisting of two blocks of 2 days each, separated by a month, in which the therapists started the treatment of two CFS patients in a pilot study. Therapists were supervised once every 2 weeks throughout the trial. Patients were allocated to therapists in a fixed sequence by the researcher in order of patients' random allocation to CBT in each centre separately. An integrity check of a random sample of 5% of all audiotaped CBT sessions was done. An independent judge used a checklist to rate the degree and the amount of time spent on the basic elements of CBT (restructuring of fatigue-related cognitions, attaining a base level of daily activity, gradual increase of physical activity, and returning to work or personal activities) in each session. The analyses showed that 91.5% of the time spent in therapy was relevant for CBT and that 87.0% of the sessions were adequate or good overall. The guided support groups were similar to CBT in terms of time spent and treatment schedule. Each group, consisting of about eight patients, had 11 meetings of 1.5 h over 8 months. One social worker was available for all 11 groups in the three centres. The treatment orientation was non-directive and client-centred. The social worker was supervised once every 2 weeks by a psychotherapist, who had no links with CBT or CFS. The goal of the support groups was to offer mutual understanding and recognition by means of exchanging experiences with one central theme during each meeting. In this study group, patients were free to have other examinations or treatments. All support-group sessions were videotaped, and the tapes were randomly checked to make sure that the social worker was not using CBT-like strategies. In the control condition natural course, no interventions were offered, and no further requirements were made. Patients were free to have other examinations or treatments. ASSESSMENTS Multidimensional assessments were made at baseline, at 8 months, and at follow-up (14 months). The baseline assessment included the screening assessment before randomisation (fatigue, functional impairment, criteria of the Centers for Disease Control and Prevention) and those made immediately after randomisation. Fatigue severity was assessed by a subscale of the checklist individual strength [14]. In this questionnaire, the patient is asked about fatigue in the 2 weeks preceding the assessment. The subscale consists of eight items, each scored on a 7-point Likert scale (range 8-56). The questionnaire has good reliability (Cronbach's a varying from 0.83 to 0.92) and discriminative validity [12, 14 and 15]. Functional impairment was measured by the sickness impact profile [16 and 17]. This widely used measure has good reliability and content validity [18]. As in our previous studies, a total score was calculated by addition of the weights of items (range 0-5799) in eight subscales: home management, mobility, alertness behaviour, sleep/rest, ambulation, social interactions, work, and recreation and pastimes. Comparison data for CFS patients were available [12]. The Karnofsky performance status scale is a descriptive, ordinal scale. An independent clinical psychologist rated the patient's functional status in 10-point intervals from 0 to 100. The validity and reliability of this scale have been shown in several populations [19 and 20]. Comparison data for CFS patients were available [6]. The symptom checklist 90[21] measured psychological wellbeing. The scale consists of 90 items scored on a 5-point Likert scale. The total score ranges from 90 to 450. A low total score reflects high psychological wellbeing. This scale is widely used, and the reliability and discriminating validity are good. The visual analogue scale of the EuroQol[22] measured quality of life. The scale ranges from 0 (worst health status) to 100 (best health status). The EuroQol has been validated in normal populations, patients, and in CFS patients [23]. Hours working in a job were recorded on a 24 h timetable of the 12-day self-observation list [24]. Self-rated improvement was measured at 8 months and at follow-up by one specific question: patients indicated whether they had completely recovered, felt much better, had the same complaints, or had become worse compared with the previous measurement. This measure has been validated in several populations of patients[24, 25 and 26] and was used in this study as one of the measures for clinically significant improvement. The self-efficacy scale, consisting of five questions, measured sense of control in relation to CFS complaints. Four items were scored on a 5-point Likert scale and one item on a 4-point Likert scale. The total score ranges from 5 to 24, a higher score reflecting more sense of control. Cronbach's a reliability coefficients range from 0.70 to 0.77 [10, 12 and 25]. Somatic attributions with respect to CFS were measured by the causal attribution list consisting of five questions scored on a 4-point Likert scale. The total score ranges from 5 to 20, a higher score indicating stronger somatic attributions. Cronbach's a reliability coefficients range from 0.71 in previous studies[12 and 25] to 0.74 in this study. Physical activity was measured by the actometer, a motion-sensing device attached to the ankle and worn continuously for 12 days. Such devices are reliable and valid measures of physical activity [27]. The activity pattern of each patient was typified by comparison of daily activity scores with the reference score of CFS patients. Three categories were defined: pervasively passive (90% or more beneath the reference score); moderately active; pervasively active (90% or more above the reference score) [28]. Focusing on bodily symptoms was measured by the subscale somatisation of the symptom checklist 90 [21], as in previous studies in which CFS patients were compared with healthy individuals and patients with multiple sclerosis [10 and 29]. The subscale consists of 12 items scored on a 5-point Likert scale. The score ranges from 0 to 60. ANALYSIS A general linear model for repeated measurements (by the method of mixed linear models) was used to analyse the effects of CBT on the two primary variables (fatigue severity and functional impairment) and the secondary variables Karnofsky performance status, symptom checklist 90, EuroQol, and hours working in a job. Differences at 8 months and 14 months from baseline were used as repeated measurements, with treatment (three levels), centre (three), time (two levels) and their first-order interactions as fixed factors. The covariance matrix was specified as unstructured (implying a general structure), the estimation method used was restricted maximum likelihood, and Satterthwaite's method was used to estimate denominator degrees of freedom. First, we tested, for the primary variables, whether the centre terms could be regarded as redundant (likelihood ratio test comparing the two models). If this was the case for both variables, reduced models with treatment and time factors and their interaction were used in all subsequent analyses. All treatment effects, as well as differences between treatments were estimated within these models; 95% CI were computed from these estimates and their standard errors. We used the procedure MIXED from the SAS package (version 6.12). Although the methods of analysis for the primary and secondary variables are the same, results for the latter should be regarded as exploratory. To define clinically significant improvement in fatigue severity, we first calculated for each patient a reliable change index to decide whether statistically significant improvement had occurred (reliable change >1.64, p<0.05). Second, a cut-off score of 36 or lower was calculated to decide whether a patient's score had moved from the range of CFS patients to the range of healthy individuals [30]. A patient was classified as showing clinically significant improvement if both criteria were met. Improvement in the Karnofsky performance status was explored also, so that we could compare the results with those of Sharpe and colleagues [6]. Clinically significant improvement was defined as an improvement of 10 points or more and a score of 80 or more. Self-rated improvement was defined as a patient's indication that he or she was completely recovered or felt much better. The categorical variables calculated by the procedures applied on the original variables checklist individual strength, Karnofsky performance status, and self-rated improvement, were compared between treatments by Fisher's exact test at 8 months and 14 months. Analyses of possible predictors were done with multiple linear regressions. The predictors were included and excluded with the stepwise method. Independent variables were treatment, baseline value of the dependent variable, age, sex, duration of complaints, education, and the baseline values of the perpetuating factors: sense of control, activity pattern, focusing on bodily symptoms, and somatic attributions, and all first-order interactions between treatment and other factors. The main interest was the relation between predictors and the direct treatment effect. Therefore, only the outcomes at 8 months were analysed. Results from these analyses should be regarded as exploratory. Results 518 patients were referred to the University Medical Centre Nijmegen with a major complaint of fatigue; CFS was diagnosed in 410. Another 66 patients were diagnosed with CFS at the University Hospital Maastricht. Of these 476 patients, 99 did not meet the eligibility criteria and 99 refused to take part. The remaining 278 patients were randomly assigned to the study groups at the centres of Nijmegen, Leiden, and Maastricht (Fig. 2). In total, 93 patients entered the CBT group, 94 the support groups, and 91 the control natural course group. Six patients were excluded: five developed other diseases during the trial and one was pregnant at baseline. After randomisation, two patients were found not to meet the criteria for CFS because they had premorbid anorexia nervosa. Thus, the trial consisted of 270 patients (92 CBT, 90 support groups, 88 control group), of whom 203 (75%) completed 8 months and 186 (69%) 14 months in the trial. 18 patients did not start treatment. 49 withdrew during the test phase and 17 withdrew during follow-up. Withdrawal was defined differently for the three groups. In the natural course group, only patients not attending the assessments were classified as withdrawing, whereas in the two intervention groups those who stopped treatment were also counted. Moreover, in contrast to CBT, frequent non-attendance in the guided support groups had no consequences for further treatment, unless a patient declared the intention to withdraw. This difference was reflected in the significant difference in mean hours of attending treatment between the CBT group and the guided support group (15.6 vs 13.2; p<0.0001). Table 1 shows the baseline characteristics of the three groups. At 8 months, 241 patients (89%) had complete data (83 CBT, 80 support groups, 78 natural course). At 14 months, 196 patients (73%) had complete data (58 CBT, 62 support groups, 76 natural course). The data of these patients were included in the analyses. Only 9% of the patients had missing data at one or both post-treatment assessments. For both primary outcome variables, a reduced model without any centre term could be used (p=0.437 for checklist individual strength, fatigue; P=0.202 for sickness impact profile, likelihood ratio test with 8 df). Consequently, all subsequent analyses were done with such models. In the primary outcome variables, significant differences between the treatment effects of CBT, support groups, and natural course were found (Fig. 3). Estimated differences are shown in Table 2. Table 3 gives the estimated differences between the study groups in secondary outcome variables. At 8 months, improvement in Karnofsky performance status, psychological wellbeing, and quality of life was statistically significantly greater in the CBT group than in either of the other groups. Differences in the time spent working in a job did not reach the 5% level of significance. Statistically significant treatment effects between CBT and support groups were found in all secondary outcome variables at 14 months. Treatment effects of CBT and natural course showed statistically significant differences for the Karnofsky performance status at both 8 months and 14 months and for psychological wellbeing at 8 months. Table 4 shows the proportions of patients with clinically significant improvements in fatigue severity, Karnofsky performance status, and self-rated improvement. For these three variables, the proportion with clinically significant improvement was statistically significantly higher in CBT than in the control conditions. All factors in the stepwise regression related to the outcome measures fatigue severity and functional impairment at p<0.05 are presented in order of entrance in the model in Table 5. The improvement in fatigue severity at 8 months was predicted by interactions of CBT with sense of control and by a passive activity pattern, rather than by CBT alone. In the CBT study group, patients with a greater sense of control at baseline had a larger decrease in fatigue severity at 8 months, immediately after CBT, than patients with lower sense of control. The reverse was true for patients with a passive activity pattern; they improved less than patients with other activity patterns. Improvement in functional impairment at 8 months was predicted by CBT alone and by interaction of CBT and focusing on bodily symptoms. Patients assigned CBT improved more than patients in both control groups. However, patients in CBT with a high level of focusing on bodily symptoms improved less than patients with lower scores on this factor. Discussion In this study, CBT was more effective for CFS patients than guided support groups or the natural course. Intention-to-treat analyses showed clinically significant improvement in fatigue severity, Karnofsky performance status, and self-rated improvement in substantial proportions of patients treated with CBT. An unexpected finding was that support groups were no more effective than the natural course (Fig. 3). This finding contrasts with other chronic diseases, in which support groups are beneficial. However, 80% or more of the patients experienced mutual understanding in the support group, and rated the contact with the therapist and the atmosphere in the group as good. These findings suggest that clinical improvement and patients' satisfaction are not correlated and may be independent. There was a large withdrawal rate in the trial, especially in the CBT and support groups. Many CFS patients eagerly expect a medical solution for their complaints and are quite sceptical about psychological treatments. Others expect more benefit from medical examinations or alternative treatments. These patients may have withdrawn prematurely. The physical burden of travelling to the centre for therapy was another reason for patients to withdraw. However, many patients who withdrew during treatment were willing to attend for assessment of the primary outcome variables. At 8 months, there was a withdrawal rate of 25%, but only 11% of the patients had missing data. Results of the analyses depend among other assumptions on that of "missingness at random", which means that missingness is possibly related to the observed data, but, conditional on these data, not to the (unknown) value of the variable itself. Although we cannot prove the assumption, we can partially check it as follows [31]: comparison of characteristics of completers and non-completers (age, sex, duration of complaints, centre, and all baseline measures) showed no differences. Furthermore, the results of intention-to-treat analyses and those of the analyses of the completers, were mostly qualitatively similar. Supporting evidence of the effectiveness of CBT was found in the significant improvement in Karnofsky performance status rated by an independent clinical psychologist in the group of patients treated with CBT compared with the control groups. A significant treatment effect on quality of life, psychological wellbeing, and work rehabilitation was only found in the comparisons of CBT with support groups and not between CBT and natural course. We were especially interested in work rehabilitation, a new element in the tested treatment protocol. The final goal of CBT for CFS included work rehabilitation for patients who used to be active in a job and resumption of other personal activities for patients without a job. We could not conclude the extent to which this goal was reached, because only hours working in a job were measured. However, in our sample of 270 patients only 33% had a job at baseline, whereas 76% had been employed before the onset of CFS. For the unemployed patients, securing employment within the limited period of treatment and follow-up would be difficult, although most of these patients did resume personal activities. The development of adequate measures of rehabilitation should have high priority in future research on CBT for CFS. The proportions of patients with clinically significant improvement in this study were lower than in other CBT trials. We suggest several explanations for this discrepancy. First, therapists in this study had no clinical experience with CFS patients at the start of the trial. Afterwards, 82% of the therapists agreed with the statement that CFS patients are more difficult to treat than patients with psychiatric diagnoses, and 54% agreed that CFS patients are more difficult to treat than patients with other functional somatic syndromes. Second, criteria for statistical and clinical significance in this study were more stringent than in previous trials. The cut-off score for clinically significant improvement was based on normative comparisons of CFS patients and healthy individuals and was perhaps overly stringent. In a recent evaluation of the concept of clinically significant improvement [32], Kendall and colleagues questioned whether patients should be compared with a non-representative `supernormal' sample of healthy people, from which all individuals with any psychological or physical disorders are excluded. Third, the treatment protocol seemed not to be suitable for a group of CFS patients who showed passive activity patterns. Analyses of prognostic factors showed that patients with this activity pattern and patients with a strong tendency to focus on bodily symptoms improved less than did patients not characterised by one of these factors. In our clinical practice, the treatment protocol has now been adjusted to both aspects. In the new treatment protocol, the emphasis is now on impeding cognitions and behaviour rather than on symptoms. Furthermore, a different treatment protocol has been developed for patients with a passive activity pattern. The early emphasis in CBT on a base level of daily activity, so important for moderately active CFS patients, seems to increase the fear of physical activity in passive CFS patients and impedes the subsequent gradual increase of physical activity. Therefore, CBT for patients with passive activity patterns starts with building up physical activity, whereas more active patients still start with attaining and maintaining a base level of daily activity. The results of this trial suggest that CBT can be transferred from CFS research clinics to therapists with no previous experience in CBT. This transfer is essential to detach the treatment from medical research settings, in which only a limited number of CFS patients can be treated. To increase accessibility of this treatment for all CFS patients in future, CBT will have to be implemented outside university medical settings. This idea accords with Wessely and colleagues' suggestion of transferring the diagnosis and treatment of functional somatic syndromes from medical subspecialists to more broadly based general physicians aided by psychiatrists or psychologists [33]. Ideally, general practitioners should diagnose CFS and refer patients to psychotherapists for CBT, without detours to medical specialists, as in other functional somatic syndromes [33]. Before this goal can be reached, expertise needs to be generalised from specialist centres to general practitioners and behaviour therapists in general (mental) health settings. CONTRIBUTORS Judith Prins coordinated data collection and analysis and drafted the report. Gijs Bleijenberg developed the original idea, was responsible for study design, study coordination, training of therapists, and development of treatment protocol, and contributed to writing of the paper. Ellen Bazelmans contributed to study design, training of therapists, and treatment protocol, supervised the therapists, and contributed to editing of the paper. Lammy Elving was responsible for the development of the diagnostic protocol and selection of patients and contributed to data collection and editing of the paper. Theo de Boo was responsible for data analysis and statistics and contributed to the writing of the paper. Johan Severens contributed to study design and editing of the paper. Gert-Jan van der Wilt contributed to the study design and data analysis. Philip Spinhoven contributed to study coordination, data collection, and editing of the paper. Jos W M van der Meer developed the original idea and contributed to study design, study coordination, and writing. Acknowledgements We thank the physicians of the departments of General Internal Medicine of the University Medical Centre Nijmegen and the University Hospital Maastricht for referring patients to the trial; the participating (assistant-) behaviour therapists and the participating research assistants of the Department of Psychotherapy, Maastricht Mental Health Institute, the Department of Psychiatry, Leiden University Medical Centre, and the Department of medical Psychology, University Medical Centre Nijmegen; the social worker for guiding support groups; and the clinical psychologist for assisting in data collection. This study was supported by a grant from the Health Insurance Council (College voor Zorgverzekeringen). Tables Table 1. Baseline characteristics of study participants(a) ----------------------------------------------------------------- CBT (n=92) Guided support Natural course groups (n=90) (n=88) ----------------------------------------------------------------- Demography Age in years 36.2 (9.4) 37.1 (10.6) 36.7 (10.3) Educational 3.9 (1.6) 4.3 (1.4) 4.4 (1.6) attainment (1=low to 7=high) M/F* 22/70 19/71 17/71 CFS features Duration in 4.9 (4.8) 6.6 (6.4) 5.3 (5.4) years CIS fatigue 52.2 (3.9) 52.3 (4.0) 51.9 (4.1) SIP total 1755 (613) 1842 (560) 1859 (671) Karnofsky 71.5 (8.5) 71.2 (7.5) 70.8 (7.9) SCL-90 170 (38.5) 169 (41.5) 166 (36.0) EuroQol 46 (17) 43 (16) 40 (14) Work, h in 12 16.3 (21.1) 12.8 (19.1) 13.5 (18.6) days Sense of 14.8 (3.5) 14.6 (3.1) 14.6 (3.6) control Somatic 13.9 (2.8) 14.1 (2.5) 13.5 (2.4) attributions Focusing on 30.7 (6.9) 30.0 (7.6) 29.8 (7.2) bodily symptoms Activity pattern** Generally 21 (23%) 16 (19%) 24 (29%) passive* Moderately 56 (62%) 53 (62%) 50 (59%) active* Generally 13 (15%) 16 (19%) 10 (12%) active* ----------------------------------------------------------------- a CIS= checklist individual strength; SIP= sickness impact profile; SCL-90= symptom checklist 90. Data are mean (SD) * number of participants. ** 11 cases had incomplete actometer data and are not included. Table 2. Estimated effect of CBT compared with support groups and natural course on fatigue severity (CIS) and functional impairment (SIP) ----------------------------------------------------------------- CBT vs support groups CBT vs natural course Treatment effect Treatment effect (95% CI) p (95% CI) p ----------------------------------------------------------------- CIS 8 months 6.0 (3.1-9.0) 0.0001 6.0 (3.1-9.0) 0.0001 14 months 5.8 (2.2-9.4) 0.0015 5.6 (2.1-9.0) 0.0016 SIP 8 months 217 (26-408) 0.0261 213 (22-403) 0.0287 14 months 263 (38-488) 0.0223 222 (3-441) 0.0470 ----------------------------------------------------------------- Table 3. Estimated effect of CBT compared with support groups and natural course on secondary endpoints ----------------------------------------------------------------- CBT vs support groups CBT vs natural course Treatment effect Treatment effect (95% CI) p (95% CI) p ----------------------------------------------------------------- Karnofsky 8 months -5.7 (-8.4 to -3.1) 0.0001 -5.2 (-7.8 to -2.6) 0.0001 14 months -6.3 (-9.6 to -3.0) 0.0002 -5.4 (-8.6 to -2.2) 0.0009 Symptom checklist 8 months 13.9 (4.3 to 23.5) 0.0048 13.4 (4.0 to 22.7) 0.0053 14 months 11.2 (1.1 to 21.3) 0.0304 6.7 (-3.0 to 16.5) 0.1767 EuroQol 8 months -7.8 (-14.0 to -1.8) 0.0114 4.0 (-10.0 to 2.0) 0.1878 14 months -9.2 (-15.6 to -2.8) 0.0049 2.3 (-8.4 to 3.8) 0.4619 Work 8 months -5.6 (-11.7 to 0.4) 0.0681 2.9 (-8.8 to 3.0) 0.3362 14 months -9.6 (-17.1 to -2.0) 0.0132 5.9 (-13.2 to 1.4) 0.1134 ----------------------------------------------------------------- Table 4. Clinically significant improvement in the treatment groups for fatigue severity, Karnofsky performance status, and self-rated improvement(a) ----------------------------------------------------------------- Number of patients with improvement p /total ----------------------------------- --------------- CBT Support Natural CBT vs CBT vs groups course support natural groups course 8 months CIS fatigue 27/83 (33%) 10/80 (13%) 10/78 (13%) 0.003 0.005 Karnofsky 29/71 (41%) 11/69 (16%) 9/75 (12%) 0.001 <0.0001 Self-rated 42/74 (57%) 12/71 (17%) 23/78 (30%) <0.0001 0.001 improvement 14 months CIS fatigue 20/58 (35%) 8/62 (13%) 13/76 (17%) 0.009 0.026 Karnofsky 28/57 (49%) 12/62 (19%) 17/75 (23%) 0.001 0.001 Self-rated 29/58 (50%) 9/62 (15%) 24/76 (32%) <0.0001 0.034 improvement ----------------------------------------------------------------- a: CIS=checklist individual strength. *Fisher's exact test. Table 5. Parameter estimates, SE, and partial R^2 of all factors related to the outcome measure fatigue severity or functional impairment (baseline minus 8 months) at p<0.05 in order of entrance to the models(a) ----------------------------------------------------------------- Factor Coefficient (SE) Partial R^2 ----------------------------------------------------------------- Fatigue severity (CIS) CBT x sense of control 0.5088 (0.0883) 0.0856 Baseline CIS 0.7010 (0.1469) 0.0515 Focusing on bodily symptoms -0.2611 (0.0838) 0.0368 CBT passive activity pattern -8.902 (2.545) 0.0208 Moderately active activity -3.439 (1.299) 0.0229 pattern Sense of control 0.3535 (0.1723) 0.0147 Sex (female) 2.761 (1.386) 0.0133 Functional impairment (SIP) Baseline SIP 0.4767 (0.0604) 0.1788 CBT 1005 (281.9) 0.0321 CBT focusing on bodily -25.06 (8.838) 0.0266 symptoms ----------------------------------------------------------------- a: CIS= checklist individual strength; SIP= sickness impact profile. Figure captions Fig. 1. Model of CFS developed and tested with LISREL (a program for the estimation of linear structural relations)[10] Fig. 2. Trial profile Fig. 3. Effect of three study conditions on the two primary outcome variables, fatigue severity (checklist individual strength [CIS]) and functional impairment (sickness impact profile [SIP]) Error BARS=SE. References 1. K. Fukuda, S.E. Straus, I. Hickie et al., The chronic fatigue syndrome: a comprehensive approach to its definition and study. Ann Intern Med 121 (1994), pp. 953-959. 2. S. Reid, T. Chalder, A. Cleare, M. Hotopf and S. Wessely, Extracts from "Clinical Evidence": chronic fatigue syndrome. BMJ 320 (2000), pp. 292-296. 3. S. Wessely, M. Hotopf and M. Sharpe. Chronic fatigue and its syndromes, Oxford University Press, Oxford (1998). 4. S. Butler, T. Chalder, M. Ron and S. Wessely, Cognitive behaviour therapy in chronic fatigue syndrome. J Neurol Neurosurg Psychiatry 54 (1991), pp. 153-158. 5. D. Bonner, M. Ron, T. Chalder, S. Butler and S. Wessely, Chronic fatigue syndrome: a follow-up study. J Neurol Neurosurg Psychiatry 57 (1994), pp. 617-621. 6. M. Sharpe, K. Hawton, S. Simkin et al., Cognitive therapy for chronic fatigue syndrome: a randomised controlled trial. BMJ 312 (1996), pp. 22-26. 7. A. Deale, T. Chalder, I. Marks and S. Wessely, Cognitive behaviour therapy for chronic fatigue syndrome: a randomized controlled trial. Am J Psychol 154 (1997), pp. 408-414. 8. V. Subramaniam, M.W. Stewart and J.F. Smith, The development and impact of a chronic pain support group: a qualitative and quantitative study. J Pain Symptom Manage 17 (1999), pp. 376-383. 9. J.H. Barlow, S.J. Macey and G.R. Struthers, Health locus of control, self-help and treatment adherence in relation to ankylosing spondylitis patients. Patient Educ Couns 20 (1993), pp. 153-166. 10. J.H.M.M. Vercoulen, C.M.A. Swanink, J.M.D. Galama et al., The persistence of fatigue in chronic fatigue syndrome and multiple sclerosis: the development of a model. J Psychosom Res 45 (1998), pp. 507-517. 11. J. Prins and G. Bleijenberg, Cognitive behavior therapy for chronic fatigue syndrome: a case study. J Behav Ther Exp Psychiatry 30 (1999), pp. 325-339. 12. J.H.M.M. Vercoulen, C.M.A. Swanink, J.F.M. Fennis, J.M.D. Galama, J.W.M. van der Meer and G. Bleijenberg, Dimensional assessment of chronic fatigue syndrome. J Psychosom Res 38 (1994), pp. 383-392. 13. K.F. Schultz, Randomised trials, human nature, and reporting guidelines. Lancet 348 (1996), pp. 596-598. 14. J.H.M.M. Vercoulen, M. Alberets and G. Bleijenberg, De Checklist Individual Strength (CIS). Gedragstherapie 32 (1999), pp. 31-36. 15. A.J.H.M. Beurskens, U. Bultmann, I.J. Kant, J.H.M.M. Vercoulen, G. Bleijenberg and G.M.H. Swaen, Fatigue amongst working people: validity of a questionnaire measure. Occup Environ Med 57 (2000), pp. 353-357. 16. H.M. Jacobs, A. Luttik, F.W.M.M. Touw-Otten and R.A. Melker, De `sickness impact profile': resultaten van een valideringsonderzoek van de Nederlandse versie. Ned Tijdschr voor Geneeskd 134 (1990), pp. 1950-1954. 17. M. Bergner, R.A. Bobbit, W.B. Carter and B.S. Gilson, The Sickness Impact Profile: development and final revision of a health status measure. Med Care 19 (1981), pp. 787-805. 18. A.F. de Bruin, L.P. de Witte, F.C.J. Stevens and J.P.M. Diederiks, The usefulness of the Sickness Impact Profile as a generic functional status measure. T Soc Gezondheidsz 70 (1992), pp. 160-170. 19. J. Schaafsma and D. Osoba, The Karnofsky peformance status scale re-examined: a cross-validation with the EORTC-C30. Qual Life Res 3 (1994), pp. 413-424. 20. M.W. O'Dell, D.P. Lubeck, P. O'Driscoll and S. Matsuno, Validity of the Karnofsky performance status in an HIV-infected sample. J Acquir Immune Defic Syndr 10 (1995), pp. 350-357. 21. W.A. Arrindell and J.H.M. Ettema. SCL-90, handleiding bij een multidimensionele psychopathologie-indicator, Swets & Zeitlinger BV, Lisse (1986). 22. EuroQol Group. EuroQol - a new facility for the measurement of health-related quality of life. Health Policy 1990; 16: 199-208. 23. C. Myers and D. Wilks, Comparison of Euroqol EQ-5D and SF-36 in patients with chronic fatigue syndrome. Qual life Res 8 (1999), pp. 9-16. 24. J.H.M.M. Vercoulen, C.M.A. Swanink, F.G. Zitman et al., A randomised double-blind, placebo-controlled study of fluoxetine in chronic fatigue syndrome. Lancet 347 (1996), pp. 858-861. 25. J.H.M.M. Vercoulen, C.M.A. Swanink, J.M.D. Galama, J.F.M. Fennis, J.W.M. van der Meer and G. Bleijenberg, Prognosis in chronic fatigue syndrome: a prospective study on the natural course. J Neurol Neurosurg Psychiatry 60 (1996), pp. 489-494. 26. A.M. van Dulmen, J.F.M. Fennis, H.G.A. Mokkink, H.G.M. van der Velden and G. Bleijenberg, Persisting improvement in complaint-related cognitions initiated during medical consultations in functional abdominal complaints. Psychol Med 27 (1997), pp. 725-729. 27. J.H.M.M. Vercoulen, E. Bazelmans, C.M.A. Swanink et al., Physical activity in the chronic fatigue syndrome: assessment and its role in fatigue. J Psychiatr Res 31 (1997), pp. 661-673. 28. S.P. van der Werf, J.B. Prins, J.H.M.M. Vercoulen, J.W.M. van der Meer and G. Bleijenberg, Identifying physical activity in chronic fatigue syndrome using actigraphic assessment. J Psychosom Res 49 (2000), pp. 372-379. 29. J.H.M.M. Vercoulen, O.R. Hommes, C.M.A. Swanink et al., The measurement of fatigue in multiple sclerosis: a comparison with patients with chronic fatigue syndrome and healthy subjects. Arch Neurol 53 (1996), pp. 642-649. 30. N.S. Jacobsen and P. Truax, Clinical significance: a statistical approach to defining meaningful change in psychotherapy research. J Consult Clin Psychol 59 (1991), pp. 12-19. 31. R.J.A. Little and D.B. Rubin. Statistical analysis with missing data, John Wiley & Sons, New York (1987). 32. P.C. Kendall, A. Marrs-Garcia, S.R. Nath and R.C. Sheldrick, Normative comparisons for the evaluation of clinical significance. J Consult Clin Psychol 67 (1999), pp. 285-299. 33. S. Wessely, C. Nimnuan and M. Sharpe, Functional somatic syndromes: one or many?. Lancet 354 (1999), pp. 936-939. -------- (c) 2001 ScienceDirect/Elsevier Science B.V.