Introduction
Chronic fatigue syndrome (CFS)
CFS is a disorder characterized by chronic, post-exertional, disabling fatigue in the absence of an alternative explanatory diagnosis (Prins et al. Reference Prins, van der Meer and Bleijenberg2006). It is sometimes called myalgic encephalomyelitis (ME), although some believe that ME is a separate condition. The prevalence of CFS is between 0.2% and 2.6% worldwide, depending on the definition used (Prins et al. Reference Prins, van der Meer and Bleijenberg2006). Prognosis is poor if untreated (Cairns & Hotopf, Reference Cairns and Hotopf2005), although specific treatments can improve outcomes (NICE, 2007).
The PACE trial was a recently reported parallel-group randomized controlled trial of patients with CFS that tested the benefits and safety of adaptive pacing therapy (APT), cognitive behavioural therapy (CBT), and graded exercise therapy (GET), when added to specialist medical care (SMC), and SMC alone (White et al. Reference White, Goldsmith, Johnson, Potts, Walwyn, DeCesare, Baber, Burgess, Clark, Cox, Bavinton, Angus, Murphy, Murphy, O'Dowd, Wilks, McCrone, Chalder and Sharpe2011). The trial demonstrated that CBT and GET can be added to SMC to safely and moderately reduce fatigue and physical disability for CFS, but that APT was not an effective addition. This is consistent with meta-analyses and systematic reviews which also suggest that both CBT and GET are moderately effective treatments (Edmonds et al. Reference Edmonds, McGuire and Price2004; Malouff et al. Reference Malouff, Thorsteinsson, Rooke, Bhullar and Schutte2008; Price et al. Reference Price, Mitchell, Tidy and Hunot2008; Castell et al. Reference Castell, Kazantis and Moss-Morris2011), something reflected in the guideline of the UK National Institute for Health and Clinical Excellence (NICE, 2007).
Pain in CFS
Pain is a common symptom in CFS that is reported by sufferers as being disabling and compromising to both physical and social function (Meeus et al. Reference Meeus, Nijs and Meirleir2007). The Centers for Disease Control (CDC) criteria for CFS contain five painful symptoms (muscle pain, joint pain, headache, sore throat, tender lymph nodes). Muscle pain is the most common of these, affecting as many as 93% of patients (Vercoulen et al. Reference Vercoulen, Swanink, Fennis, Galama, van der Meer and Bleijenberg1994; King & Jason, Reference King and Jason2005). Chronic idiopathic pain, such as that which occurs in CFS, is poorly understood. Management strategies for chronic pain often include CBT and GET in combination with SMC, although with variable success (Morley et al. Reference Morley, Eccleston and Williams1999; Andrews et al. Reference Andrews, Strong and Meredith2012). Although the UK National Institute for Health and Clinical Excellence (NICE) recommend CBT and GET for CFS (NICE, 2007), no guidance specific to the management of pain in CFS currently exists and research in this area is sparse. It remains unknown whether CBT and GET individually combined with SMC alleviate pain as well as fatigue in CFS. It has been suggested that CBT is effective in reducing painful symptoms in CFS by one group (Knoop et al. Reference Knoop, Stulemeijer, Prins, van der Meer and Bleijenberg2007). However, this was examined in those who had recovered from CFS after successful treatment with this intervention, compared to those who had received CBT for CFS but had not recovered. This data alone is unable to clarify whether CBT is effective for painful symptoms as a consequence, or independent, of its effect on fatigue.
An overview of the precise use of therapies in the PACE trial has been described elsewhere (White et al. Reference White, Goldsmith, Johnson, Potts, Walwyn, DeCesare, Baber, Burgess, Clark, Cox, Bavinton, Angus, Murphy, Murphy, O'Dowd, Wilks, McCrone, Chalder and Sharpe2011). Briefly, CBT performed in the PACE trial was employed on the basis of the fear avoidance theory of CFS, suggesting that the cognitive (fear of engaging in activity) and behavioural (avoidance of activity) responses are linked and interact at the physiological level to perpetuate fatigue. The aim of treatment was to target these responses. GET performed in the PACE trial was employed on the basis of deconditioning and exercise intolerance theories of CFS. The aim of treatment was to help the participant gradually to habituate to appropriate physical activities, reverse the deconditioning, and thereby reduce fatigue and disability. The theories underpinning the use of CBT and GET in CFS for the PACE trial are also applicable to their use in the management of chronic pain (Ostelo et al. Reference Ostelo, van Tulder, Vlaeyen, Linton, Morley and Assendelft2005; Lohnberg, Reference Lohnberg2007). APT was designed to help the patient better adapt to their illness by listening to their body and balancing activity against rest.
The aim of our study was to use data from the PACE trial to compare treatments for their effect on pain, reported by people with CFS, and to determine whether treatment responses were associated with changes in fatigue.
We hypothesized that CBT and GET would be more effective than SMC or APT for painful symptoms based upon prior literature and the evidence that these treatments were more effective for reducing both fatigue and disability (White et al. Reference White, Goldsmith, Johnson, Potts, Walwyn, DeCesare, Baber, Burgess, Clark, Cox, Bavinton, Angus, Murphy, Murphy, O'Dowd, Wilks, McCrone, Chalder and Sharpe2011).
Methods
We conducted an analysis of the PACE trial data, comparing the effects of APT, CBT, GET, and SMC on pain and co-morbid fibromyalgia.
In total, 641 patients were recruited into the PACE trial from secondary-care specialist CFS clinics in the UK, all of whom met the Oxford criteria for CFS (6 months of disability, with fatigue as a principal symptom, not otherwise explained by other diagnoses) (Sharpe et al. Reference Sharpe, Archard, Banatvala, Borysiewicz, Clare, David, Edwards, Hawton, Lambert and Lane1991). Participants were assessed by a research assistant for the presence of the international CDC criteria for CFS (Reeves et al. Reference Reeves, Lloyd, Vernon, Klimas, Jason, Bleijenberg, Evengard, White, Nisenbaum and Unger2003) and diagnostic criteria for fibromyalgia (Wolfe et al. Reference Wolfe, Clauw, Fitzcharles, Goldenberg, Katz, Mease, Russell, Russell, Winfield and Yunus2010). The international criteria for CFS require 6 months of disabling fatigue and at least four of eight associated symptoms (which include five pain symptoms), not otherwise explained by an alternative diagnosis (Reeves et al. Reference Reeves, Lloyd, Vernon, Klimas, Jason, Bleijenberg, Evengard, White, Nisenbaum and Unger2003). We used the American College of Rheumatology criteria (Wolfe et al. Reference Wolfe, Clauw, Fitzcharles, Goldenberg, Katz, Mease, Russell, Russell, Winfield and Yunus2010) to diagnose fibromyalgia, with the exception of tender points. These criteria require having 3 months of widespread pain in all four quadrants, above and below the waist, on both sides of the body, as well as axial pain (Wolfe et al. Reference Wolfe, Clauw, Fitzcharles, Goldenberg, Katz, Mease, Russell, Russell, Winfield and Yunus2010). All measures and criteria were assessed at baseline (week 0, trial entry) and at the study endpoint (week 52). The methodology, including power calculations and descriptions of the interventions performed and primary outcome results of the PACE trial have been reported in detail elsewhere (White et al. Reference White, Sharpe, Chalder, DeCesare and Walwyn2007, Reference White, Goldsmith, Johnson, Potts, Walwyn, DeCesare, Baber, Burgess, Clark, Cox, Bavinton, Angus, Murphy, Murphy, O'Dowd, Wilks, McCrone, Chalder and Sharpe2011).
Painful symptom outcomes
While all patients in the PACE trial were required to meet the Oxford criteria for CFS for inclusion, data was also collected for the international (CDC) criteria. We have focused on these in the current analysis due to their description of painful symptoms. The international (CDC) criteria for CFS contain five painful symptoms central to a diagnosis of CFS: muscle pain, joint pain, headache, sore throat and tender lymph nodes. We chose muscle pain and joint pain as individual symptom outcomes, before analysis of outcome data, since these were the most commonly reported at baseline. Each of the symptom domains for the CDC checklist was scored with a 5-point scale: not at all present (score 0), present a little (1), present more than not (2), present most of the time (3), present all of the time (4).
A further categorical scoring system was generated for both CFS and non-CFS pain, based upon presence (1) or absence (0) of symptoms. For the CFS-related symptoms (CDC criteria), ratings of 0 and 1 were scored as ‘absent’ and ratings of 2, 3 or 4 were rated as ‘present’.
We used the frequency scores for our comparisons of individual pain symptoms. This was a tertiary outcome chosen prior to the present analysis but after the main analysis of the PACE trial, which did not include painful symptoms. Diagnostic criteria for fibromyalgia were recorded as present (1) or absent (0) at baseline (study entry) and study endpoint (52 weeks) for all participants. Fibromyalgia is commonly co-morbid with CFS and since there is also a clinical overlap between the two disorders, in terms of painful symptoms, it was felt important to include this data in the analysis.
Analysis
Independent-sample t tests were used to compare the unadjusted means of pain frequency scores for individual and global pains between individual treatment arms, comparing two at a time (CBT v. APT, CBT v. GET, CBT v. SMC, APT v. GET, APT v. SMC, SMC v. GET). The PACE trial was not powered to investigate CBT v. GET (White et al, Reference White, Sharpe, Chalder, DeCesare and Walwyn2007). Pearson's χ 2 test was used to compare categorical data between trial arms for the presence or absence of the diagnostic criteria for fibromyalgia. We did not adjust for multiple comparisons because the pain outcomes were secondary in the PACE trial, and the five comparisons were stated a priori (White et al. Reference White, Sharpe, Chalder, DeCesare and Walwyn2007).
We calculated effect sizes for the unadjusted, significant differences between treatment arms by Cohen's d (Cohen, Reference Cohen1988) at 52 weeks.
To adjust for confounders, composite CFS pain scores and the most common CFS pain domains (muscle pain and joint pain) were individually modelled by multiple linear regression with baseline depressive disorder, international (CDC) criteria for CFS, and London ME criteria, in addition to trial arm and the dependent pain score at baseline. The presence of fibromyalgia diagnostic criteria was modelled using binary logistic regression using the same covariates and the presence of the diagnostic criteria at baseline. All analyses were performed using the Statistical Package for the Social Sciences (SPSS, IBM, version 20, USA). Confidence intervals were calculated using Wilson's method (Bryant, Reference Bryant, Altman, Machin, Bryant and Gardner2000).
Results
A total of 641 patients were recruited, one of whom withdrew consent after participation. Demographic and clinical characteristics of participants were similar across treatment arms, apart from a shorter duration of illness in the SMC group (Table 1). Further details of the sample have been reported previously (White et al. Reference White, Goldsmith, Johnson, Potts, Walwyn, DeCesare, Baber, Burgess, Clark, Cox, Bavinton, Angus, Murphy, Murphy, O'Dowd, Wilks, McCrone, Chalder and Sharpe2011). The frequencies of the five individual pain symptoms were 70% for muscle pain, 51% for joint pain, 37% for headache, 31% for sore throat and 31% for lymph node pain.
Table 1. Selected demographic and clinical characteristics at baseline, n (%)
APT, Adaptive pacing therapy; CBT, cognitive behavioural therapy; GET, graded exercise therapy; SMC, specialist medical care; CFS, chronic fatigue syndrome; ME, myalgic encephalomyelitis; s.d., standard deviation.
Descriptive statistics
Frequencies of all painful symptoms and the presence of fibromyalgia diagnostic criteria are presented in Tables 2 and 3.
Table 2. Presence of individual symptoms, valid% (n)
APT, Adaptive pacing therapy; CBT, cognitive behavioural therapy; GET, graded exercise therapy; SMC, specialist medical care;
Table 3. Presence of fibromyalgia diagnostic criteria; valid% (n) and unadjusted trial arm comparisons,% difference (95% CI), p value
APT, adaptive pacing therapy; CBT, cognitive behavioural therapy; GET, graded exercise therapy; SMC, specialist medical care; CI, confidence interval.
* p < 0.05.
Unadjusted comparisons
Pain symptoms
Individual pain symptoms are given in Table 4. Both CBT and GET were associated with less frequent muscle pain compared to APT (p = 0.04 and 0.03, respectively), and with SMC (p = 0.02 and 0.01, respectively). Both CBT and GET were associated with less frequent joint pain compared to APT (p = 0.04 and 0.03, respectively), but not compared to SMC.
Table 4. Trial arm comparisons for the most frequently occurring pain symptoms at 52 weeks – reduced score implies reduction in symptoms
APT, adaptive pacing therapy; CBT, cognitive behavioural therapy; GET, graded exercise therapy; SMC, specialist medical care; CI, confidence interval.
* p < 0.05.
The calculated effect sizes for these significant differences between treatment arms varied between 0.25 and 0.31 for muscle pain and 0.24 and 0.26 for joint pain.
Fibromyalgia
GET was associated with a significantly greater reduction in the frequency of fibromyalgia at 52 weeks, compared to APT (p = 0.04) and SMC (p = 0.01) (Table 3). The number needed to treat (NNT) for these comparisons were 11 and 13, respectively.
Regression modeling
Muscle pain
After adjusting for possible confounders using linear regression, reductions in muscle pain after both CBT (difference = 0.14, p = 0.05) and GET (difference = 0.38, p = 0.01) remained significantly greater than with SMC, and GET remained significantly greater than APT (difference 0.17, p = 0.01). The difference between CBT and APT was confounded by baseline differences in muscle pain and was no longer significant (p = 0.12). Adding change in fatigue score into the model had little effect on the treatment comparisons (CBT v. SMC: 0.14, p = 0.05; GET v. APT: 0.18, p = 0.01; GET v. SMC: 0.42, p < 0.001). Participant subgroups meeting international criteria for CFS, London criteria for ME, and depressive disorder criteria did not differ in the pattern of treatment effects.
Joint pain
The effect of both CBT (difference = 0.35, p = 0.02) and GET (difference = 0.15, p = 0.02) on joint pain was significantly greater than that of APT after adjusting for baseline confounders. There was a trend for both CBT (difference = 0.14, p = 0.05) and GET (difference = 0.22, p = 0.07) to be associated with less joint pain compared to SMC, which was not evident in the unadjusted comparison (p = 0.14 and 0.12, respectively). Adding change in fatigue into the model had little effect on treatment comparisons, with the differences barely changing for CFS joint pain (CBT v. APT: 0.31, p = 0.04; CBT v. SMC: 0.13, p = 0.08; GET v. APT: 0.15, p = 0.03; GET v. SMC: 0.25, p = 0.05). As before, alternative diagnostic criteria and the presence of major depressive disorder did not alter the pattern of treatment effects.
Fibromyalgia diagnostic criteria
A statistically significantly greater reduction in the frequency of fibromyalgia at 52 weeks was seen for GET compared to SMC in the adjusted model (difference = 0.73, p = 0.03). Adding change in fatigue did not alter the model, with the difference barely changing (GET v. SMC: 0.70, p = 0.05). As before, alternative diagnostic criteria and the presence of major depressive disorder did not alter the pattern of treatment effects.
Discussion
This study found that GET was more effective than either APT or SMC in reducing the frequencies of both muscle and joint pain. It was also more effective than SMC in reducing the prevalence of fibromyalgia. CBT was more effective than SMC in reducing the frequency of muscle pain and more effective than either APT or SMC in reducing the frequency of joint pain. The beneficial effects of these therapies on painful symptoms in CFS were independent of their effects on reducing fatigue, i.e. their effects on painful symptoms were not dependent on generating an improvement in fatigue. However, the effect sizes were small.
Limitations
The timings of data collection during the PACE trial were such that we were unable to report a linear trend in the response to the interventions. Our measures of pain were based on frequencies of pain, rather than severity of pain and as such may not be as clinically meaningful. By virtue of the original power calculations for the PACE trial, we are unable to report on comparisons between the two most effective interventions, CBT and GET. The PACE trial also excluded patients unable to attend hospital, potentially excluding the more severely affected patients, who may have more pain that is less responsive, or differently responsive to the interventions assessed here. Similarly, the group of patients assessed by the PACE trial were those referred to secondary care, rather than those managed within general practice. These factors may affect the generalizability of our findings.
Strengths
This analysis has the benefit of the strengths of the original PACE trial data, which had only small numbers of dropouts (5%), high rates of compliance with the treatments, use of manual-defined treatments provided by competent therapists, with good therapeutic alliance and high rates of participant satisfaction (White et al. Reference White, Goldsmith, Johnson, Potts, Walwyn, DeCesare, Baber, Burgess, Clark, Cox, Bavinton, Angus, Murphy, Murphy, O'Dowd, Wilks, McCrone, Chalder and Sharpe2011). Furthermore, we examined all trial participants by trial arm, irrespective of primary outcome, enabling an analysis of the effect of interventions on painful symptoms in CFS rather than an effect on painful symptoms that was dependent on the presence of or reduction of fatigue (Knoop et al. Reference Knoop, Stulemeijer, Prins, van der Meer and Bleijenberg2007).
We additionally examined the effects of treatments on fibromyalgia and demonstrated that GET directed at CFS improves outcome for co-morbid fibromyalgia as well as fatigue. CFS and fibromyalgia are similar disorders and commonly co-morbid (McKay et al. Reference McKay, Duffy and Martin2009). That GET is effective when the two co-exist has important clinical ramifications. However, the NNT for a positive outcome is greater than that reported for ‘trial recovery’ after 52 weeks of either GET or CBT in the PACE trial (White et al. Reference White, Goldsmith, Johnson, Chalder and Sharpe2013).
Comparisons with previous work in this area
CBT in CFS focuses on changing maintaining factors such as unhelpful avoidance in order to reduce fatigue and associated disabilities (Whiting et al. Reference Whiting, Bagnall, Sowden, Cornell, Mulrow and Ramirez2001). Although trials examining the effects of CBT in CFS report positive treatment effects, fatigue is normally the primary outcome and pain is rarely measured (Chambers et al. Reference Chambers, Bagnall, Hempel and Forbes2006). Two previous studies have shown improved pain outcomes after CBT for CFS (Stulemeijer et al. Reference Stulemeijer, de Jong, Fiselier, Hoogveld and Bleijenberg2005; Knoop et al. Reference Knoop, Stulemeijer, Prins, van der Meer and Bleijenberg2007). A randomized controlled trial of CBT for CFS in adolescents reported reduced muscle pain and headaches in addition to reduction in fatigue (Stulemeijer et al. Reference Stulemeijer, de Jong, Fiselier, Hoogveld and Bleijenberg2005). More recently Knoop and colleagues looked at adult patients with CFS who had recovered from fatigue versus those with CFS who had not recovered from fatigue after a CBT intervention for CFS (Knoop et al. Reference Knoop, Stulemeijer, Prins, van der Meer and Bleijenberg2007). They found that recovered patients reported a more significant reduction in pain and fewer pain locations compared to the non-recovered patients. The sampling of patients by the efficacy of an intervention for one outcome (fatigue) to examine the effect on another outcome (pain) is likely to be biased towards the intervention if there is an association between the two outcomes, and should therefore be interpreted with caution. This is consistent with the contrasting findings of this trial, which found that efficacy against pain was independent of the effect on fatigue.
Both CBT and GET are also commonly used in the management of other chronic idiopathic pain disorders although success in this regard seems to vary between treatments and disorders (Morley et al. Reference Morley, Eccleston and Williams1999; Andrews et al. Reference Andrews, Strong and Meredith2012). CBT is associated with greater effects for any psychological interventions used (Glombiewski et al. Reference Glombiewski, Sawyer, Gutermann, Koenig, Rief and Hofmann2010). It is also effective in altering positive coping measures and reducing the behavioural expression of pain but is less effective against catastrophization, affective disturbance and social functioning as they relate to pain (Morley et al. Reference Morley, Eccleston and Williams1999). This may lead to the hypothesis that GET might be superior to CBT in the treatment of chronic idiopathic pain. However, in the current analysis we are unable to report on this, as the PACE trial was not powered to examine this comparison directly (White et al. Reference White, Sharpe, Chalder, DeCesare and Walwyn2007, Reference White, Goldsmith, Johnson, Potts, Walwyn, DeCesare, Baber, Burgess, Clark, Cox, Bavinton, Angus, Murphy, Murphy, O'Dowd, Wilks, McCrone, Chalder and Sharpe2011).
We found significant advantages for both CBT and GET for the most common types of pain (particularly those affecting muscles). Joint pain was significantly better relieved by CBT compared to APT, while adjusted comparisons showed only a trend for CBT to be better than SMC. This may be related to the encouragement of self-help management incorporated within SMC.
GET was the most effective intervention for pain in this trial, with GET reducing muscle pain more than APT and SMC, reducing joint pain more than APT, and co-morbid fibromyalgia more than SMC. Physical activity (as compared to sustained sedentary behaviour) appears important in the central nervous system's regulation of pain in fibromyalgia, a condition closely related to and commonly co-morbid with CFS (Ellingson et al. Reference Ellingson, Shields, Stegner and Cook2012). GET is a moderately effective treatment for several pain disorders beyond CFS (Geraets et al. Reference Geraets, Goossens, de Groot, de Bruijn, de Bie, Dinant, van der Heijden and van den Heuvel2005; George et al. Reference George, Wittmer, Fillingim and Robinson2010). The differences in approach between APT and GET highlight the importance of a graded approach to increasing physical activity for improving outcomes for chronic pain.
APT has been used in other chronic idiopathic pain but recent studies report that it is ineffective (Kindermans et al. Reference Kindermans, Huijnen, Goossens, Roelofs, Verbunt and Vlaeyen2011; Andrews et al. Reference Andrews, Strong and Meredith2012). The suggestion has been made that APT represents a combination of pain avoidance and activity avoidance, both of which are independently associated with worse pain outcomes and greater disability due to pain (Andrews et al. 2011). This might explain why our analysis was unable to demonstrate any beneficial effect of APT on pain in CFS and the same may be true of its lack of effect on fatigue as an outcome in CFS, as reported in the PACE trial (White et al. Reference White, Goldsmith, Johnson, Potts, Walwyn, DeCesare, Baber, Burgess, Clark, Cox, Bavinton, Angus, Murphy, Murphy, O'Dowd, Wilks, McCrone, Chalder and Sharpe2011).
Conclusions
Pain in CFS remains poorly understood. In spite of this, both CBT and GET show some efficacy compared to more passive forms of rehabilitation. It is important that GET directed at fatigue in CFS is also effective against co-morbid fibromyalgia, particularly given the frequency with which the two disorders co-occur. Advances in treatment may come from focusing these therapies towards pain as well as obtaining an improved understanding of the pathophysiology of pain in CFS.
Acknowledgements
The funders of the PACE trial were: UK Medical Research Council, Department of Health for England, Department for Work and Pensions, Scottish Chief Scientist Office. The writing of this paper and its analysis were funded by Barts and The London Charity. Trial registration: ISRCTN54 285 094. The PACE trial management group consisted of the authors of this paper as well as (in alphabetical order): Brian J. Angus, Hannah Baber, Jessica Bavinton, Mary Burgess, Julia C. DeCesare, Lucy V. Clark, Diane L. Cox, Eleanor Feldman, Paul McCrone, Gabrielle Murphy, Maurice Murphy, Hazel O'Dowd, Tim Peto, Laura Potts, Rebecca Walwyn, and David Wilks.
Declaration of Interest
P.D.W. has done voluntary and paid consultancy work for the United Kingdom government and a reinsurance company. T.C. has received royalties from Sheldon Press and Constable and Robinson. M.S. has done voluntary and paid consultancy work for the United Kingdom government, has done consultancy work for an insurance company, and has received royalties from Oxford University Press.
