Background
Description of the condition
The 1990 American College of Rheumatology (ACR) criteria for classification of fibromyalgia (FM) syndrome define it as widespread pain for longer than three months duration, with pain on palpation of at least 11 of 18 specified tender points on the body (Wolfe 1990). While the ACR criteria are most frequently used in research studies, clinicians may also employ the American Pain Society criteria for a clinical diagnosis of FM (Burckhardt 2005), that include the presence of widespread pain (all four quadrants of the body and along the midline axial) for at least three consecutive months and pain on palpation of 9 of 11 bilateral sites on the body. A 1996 consensus report offers a broader picture of FM, describing it as a "syndrome of widespread pain, decreased pain threshold, and characteristic symptoms that include non-restorative sleep, fatigue, stiffness, mood disturbance, irritable bowel syndrome, headache, paresthesias, and other less common features" (Wolfe 1996, page 534). The American Pain Society Guideline for Management of Fibromyalgia Syndrome in Adults and Children (Burckhardt 2005) also acknowledge that other symptoms including fatigue, headache, poor sleep, psychological distress and cognitive dysfunction often are part of the syndrome and can have a substantial impact on an individual's physical and emotional function and overall health-related quality of life.
Wolfe 1995 reported the prevalence of FM (all ages) to be 2% (females 3.4%, males 0.5%). A recent large-scale Canadian study (McNalley 2006) describes self-reported prevalence of FM as 1.1% for all ages (1.83% in females, 0.33% in males) with a female to male ratio of six to one. Limitations in activities associated with daily living have been reported to be as high in FM patients as in patients with rheumatoid arthritis (Hawley 1991). In individuals who seek medical attention, the condition is chronic and non-remitting, with symptoms affecting every aspect of life, including work, family life and leisure (Henriksson 1994). Researchers have reported a substantial impact of FM on ability to work and productivity. Twenty to 50% of persons with FM could work few or no days (Ledingham 1993, Wolfe 1997), 36% had an average of two or more absences from work per month (Martinez 1995), and 26.5% to 55% had received disability or social security payments (Martinez 1995, Wolfe 1997).
Many individuals with FM have been shown to be sedentary (Clark 1993) and with levels of cardiorespiratory fitness well below average (Bennett 1989, Burckhardt 1989, Clark 1993, Clark 1994). While the underlying pain, fatigue and depression are likely to contribute to sedentary lifestyles and therefore low levels of fitness, the studies being evaluated indicate that individuals with FM are able to perform maximal tests cardiorespiratory fitness, low and moderate intensity aerobic exercise, flexibility and muscle strengthening exercise.
Description of the intervention
Despite examination of a wide range of treatments, optimal management of FM is still unknown. Current reviews (Bellamy 1998, Berman 1999, Burckhardt 2002, Hadhazy 2000, Holdcraft 2003, Mannerkorpi 2003, Sim 2002) and evidence-based guidelines (Goldenberg 2004) have examined a range of treatment options divided into pharmacologic and non-pharmacologic. Non-pharmacologic strategies include interventions classified as mind-body cognitive/cognitive-behavioral, exercise, complementary and alternative therapies. Goldenberg 2004 concluded that "despite the chronicity and complexity of FM, there are pharmacological and non-pharmacological interventions available that have clinical benefit. Based on current evidence, a stepwise program emphasizing education, certain medications, exercise, cognitive therapy, or all 4 should be recommended" (page 2388). Goldenberg 2004 goes on to advise that optimal management is "best arrived at when patients and health care professionals work as a team" (page 2394). However, while exercise is recognized as one part of the management of FM, not all of the clinically relevant and practically important aspects of an exercise prescription have been elucidated.
How the intervention might work
While pain in individuals with FM may be related to central nervous system pain processing abnormalities that include central sensitization and inadequate pain inhibition, peripheral tissues including muscle may contribute to chronic pain through initiating and/or maintaining central sensitization (Staud 2005, Staud 2006). Exercise may thus contribute to pain through the process of muscle microtrauma, repair and adaptation associated with normal acute exercise and exercise training. Several studies have described metabolic findings in muscle tissue that are consistent with deconditioning (Bennett 1989, Elvin 2006, Lund 1986, Park 1998, Bengtsson 1986a, Bengtsson 1986b,Jubrias 1994). The metabolic adaptations induced by aerobic and by strength training may normalize some of these findings, thus contributing to improvements in pain (Costill 1979, Deschenes 2002, Holloszy 1984).
Exercise training has been used successfully to address a number of conditions that are also commonly experienced by individuals with FM. Aerobic and strength training have been shown to improve depression in individuals with clinical depression (Brosse 2002, Dunn 2001). Moderate exercise can improve sleep in individuals with sleep complaints (King 1997, Singh 1997). One can also reflect on training-induced improvements in cardiorespiratory fitness to suggest that fatigue may also improve because as one's maximal aerobic capacity improves, the individual will be performing activities of daily living at lower absolute percentages of maximal capacity.
Why it is important to do this review?
Incorporating exercise into one's weekly routine is not a small endeavour. It is the responsibility of clinicians and researchers to identify for clients with FM both the effects that they can expect in terms of FM signs and symptoms and the most efficacious methods of achieving those effects. This review is necessary to determine the effectiveness of various types and training volumes of exercise for improvement of FM signs and symptoms. The review should also examine what outcomes are most impacted by exercise in this population. The review is also needed to guide clinicians and individuals with FM through the maze of studies towards the currently known best prescriptions for, and ways to perform exercise by individuals with FM.
Objectives
The primary objective of this systematic review was to evaluate the effects of exercise training including cardiorespiratory (aerobic), muscle strengthening, and/or flexibility exercise on global well-being, selected signs and symptoms, and physical function in individuals with FM.
Methods
Criteria for considering studies for this review
Types of studies
We selected randomized clinical trials (RCT) that compared an intervention that included an exercise component with an untreated control or a non-exercise intervention. Studies were included if the authors used words such as randomly, random or randomization, to describe the method of assignment of subjects to groups.
Types of participants
The studies used a variety of published criteria for the diagnosis of FM: Smythe 1981, Wolfe 1990, Yunus 1981,Yunus 1982, Yunus 1984. Although some differences exist between the diagnostic criteria, for the purpose of this review all were considered to be acceptable and comparable. While exclusion criteria varied among studies, all allowed for exclusion of individuals with medical conditions for which exercise could be either contraindicated or unsafe under unmonitored conditions.
Types of interventions
Exercise was defined as the "planned, structured and repetitive bodily movement done to improve or maintain one or more components of physical fitness" (ACSM 2001). In classifying exercise interventions, we recognized two types of interventions. Composite interventions included both an exercise and non-exercise component(s) delivered simultaneously. Exercise-only interventions did not include a non-exercise component and were classified by the predominant type exercise. In classifying the type of exercise, the exercise performed in warm-up and cool-down were not considered. Exercise-only interventions included aerobic-only training, strength-only training, flexibility-only training, or mixed exercise-only interventions. No restrictions on frequency, intensity or duration were made beyond requiring that the exercise component of composite interventions be a substantial part of that treatment.
Types of outcome measures
Outcome measures did not form part of the inclusion criteria for this review. We grouped the outcome measures into six constructs representing global well-being, commonly experienced signs and symptoms of FM and observer-measured physical function.
Primary outcomes
Primary outcomes represented four constructs.
1. Pain (e.g., visual analogue scale (VAS) or 10 point ordinal scale)
2. Global well-being (overall feeling of well-being) or perceived improvement in FM symptoms (outcomes such as the Fibromyalgia Impact Questionnaire (FIQ) total score, study participant-rated change in FM symptoms, observer-rated change in FM symptoms). (Note: The FIQ is a self-report questionnaire developed to evaluate overall impact of FM. The individual with FM rates commonly experienced FM symptoms, including (but not restricted to) pain, fatigue, depression, anxiety, level of restedness after sleep, and effect of FM on work. The scores for each item can be reported individually or summed to report the FIQ total score. In this review we have used the FIQ total score to represent overall or global well being.)
3. Physical function
a. Physical performance -aerobic (e.g., submaximal or maximal treadmill or cycle ergometer tests, 6 minute walk)
b. Physical performance -musculoskeletal (e.g., grip strength, hip and knee extension strength)
c. Physical performance- flexibility (e.g., sit and reach test)
4.Tender points (e.g., pain threshold of tender points using dolorimetry or tenderness to thumb pressure)
Secondary outcomes
Secondary outcomes represented two constructs.
5. Depression (e.g., Beck Depression Inventory, FIQ subscale for depression).
6. Fatigue and sleep (e.g., FIQ fatigue subscale, sleep VAS)
Search methods for identification of studies
In the original review we searched MEDLINE (1966-12/2000), CINAHL (1982-12/2000), HealthSTAR (1990-12/2000), SportDiscus (1975 - 12/2000), EMBASE (1974 to 12/2000), and the Cochrane Controlled Trials Register (CCTR, Issue 4, 2000).
Electronic searches
The search terms and parameter of our earlier review are provided in Table 1. For this update, we searched the following databases from 1/2000 to 5/2005: MEDLINE (Ovid), EMBASE (Ovid), CINAHL (Ovid), SportDiscus (Ovid), PubMed, PEDro, and the Cochrane Central Register of Controlled Trials Register (CENTRAL, Issue 3, 2005). The primary search terms were:
1. exp FIBROMYALGIA/
2. fibromyalgia.tw.
3. fibrositis.tw.
4. or/1-3
5. exp exercise/
6. exp EXERTION/
7. exp Physical Fitness/
8. exp Exercise Test/
9. exp Exercise Tolerance/
10. exp SPORTS/
11. exp PLIABILITY/
12. exp Physical Endurance/
13. exertion$.tw.
14. exercis$.tw.
15. sport$.tw.
16. ((physical or motion) adj5 (fitness or therapy or therapies)).tw.
17. (physical$ adj2 endur$).tw.
18. manipulat$.tw.
19. (skate$ or skating).tw.
20. jog$.tw.
21. swim$.tw.
22. bicycl$.tw.
23. (cycle$ or cycling).tw.
24. walk$.tw.
25. (row or rows or rowing).tw.
26. weight train$.tw.
27. muscle strength$.tw.
28. or/5-27
29. 4 and 28
30. limit 29 to yr=2000 - 2005
Other sources
Reference lists from identified articles, meta-analyses and reviews of all types of treatment for FM were reviewed independently by two reviewers and all promising references were scrutinized.
Data collection and analysis
Selection of studies
Two reviewers (AJB, CLS) independently scanned the titles and reviewed abstracts of studies generated from searches. The reference lists from bibliographies of review articles were also examined and abstracts were retrieved for all promising titles. We retrieved all complete publications for the promising abstracts. The full text articles were then examined independently by two reviewers to determine if they met the selection criteria. Disagreements between the two reviewers were resolved in consensus meetings of the full team. Foreign language studies were translated and included in the review. In our opinion, no important papers were missed.
Data extraction and management
For the preparation of the first review (16 studies, Busch 2002), two reviewers (AJB, CLS) independently extracted data (study characteristics, study results). Point estimates for selected variables were extracted by one of the reviewers and a research assistant, and checked by a pair of reviewers. For the preparation of this major update, two of four reviewers (AJB, CLS, KARB, TJO) independently extracted data for each study. All discrepancies were rechecked and consensus achieved by discussion.
Assessment of methodological quality of included studies
Two instruments for assessing methodological quality were applied in this review: the van Tulder Methodological Quality Criteria (van Tulder 1997, van Tulder 2003) and the Jadad Methodological Quality Criteria (Jadad 1996) The van Tulder Methodological Quality Criteria were applied with two deviations from those of van Tulder 1997. We interpreted 'patient blinding' to mean rigorous information control because it is not possible to blind subjects to an exercise intervention (item h). We used a withdrawal rate of 20% (item l) as acceptable and awarded positive scores if data from at least 80% of subjects were analysed at completion of the primary short-term end point of the study, or if all subjects who entered the study were analysed at completion (i.e., intention-to-treat analysis). The Jadad Methodological Quality Criteria were applied exactly as described by Jadad 1996.
Before beginning the methodological evaluation for the 2002 review, three reviewers (AJB, CLS, PMP) independently evaluated a sample of two studies and subsequently agreed upon a consistent interpretation of criteria for each of the two instruments. The three reviewers then independently applied the two instruments, using standardized forms for each. Differences in ratings were resolved by consensus. For the current update, reviewers (AJB, CLS, KARB, TJO) worked in pairs, independently evaluating a sample of two studies and then reviewing the established interpretation of criteria for the instruments. Consensus on further clarification of the interpretation was achieved in a meeting of the four reviewers. Reviewers then worked in pairs, independently applying the two instruments, using standardized forms for each. Differences in ratings were resolved by consensus. To avoid bias, one of the included studies (Schachter 2003) which was authored by three of the current reviewers was examined by the two reviewers not involved in that study. Inter-rater reliability calculated for the updated assessment using Kappa was very good (K = 0.914, "almost perfect" according to Landis 1977).
In this update we consider the 11 items of the van Tulder methodological criteria that reflect internal validity (van Tulder 2003). We arbitrarily classified studies into high, moderate and low quality studies based on scores on these eleven items. Studies achieving a score of 8-11 were classified as being of high quality; studies scoring 5-7 were classified as being of moderate quality and those scoring 1-4 were classified as low quality studies. In this review, we place greater weight on moderate to high quality studies (i.e., those with a score of 5 or greater). In actuality, this represented a score of 50% or greater because one of the 11 items (the blinding of the care provider) is seldom achieved in exercise studies.
Evaluation of congruence of exercise/physical activity with recognized guidelines
We used the American College of Sport Medicine (ACSM) guidelines (ACSM 2001; ACSM 2006), to evaluate whether interventions had provided a training stimulus that would effect changes in physical fitness. The ACSM recommendations for achieving improvements in physical fitness represent widely accepted criteria. Since exercise guidelines have not been developed for those with FM, the ACSM guidelines (developed for healthy individuals) were used.
ACSM guidelines:(ACSM 2006)
1. Cardiorespiratory Endurance (Aerobic Training) The dosage required is as follows: a) frequency of exercise at least three days per week, b) intensity of exercise sufficient to achieve equal to or greater than 40% of heart rate reserve (min-max: 40-85%) or 64% of predicted maximum heart rate (min-max: 64-94%), c) sessions of at least 20 minutes duration (min-max: 20-60 minutes), either as continuous exercise or spread intermittently throughout the day in blocks of 10 minutes or more, and using any mode of aerobic exercise involving use of major muscle groups in rhythmic activities, d) for a total time period of at least six weeks. While ACSM recommends an exercise frequency of 3-5 days per week, it acknowledges that "deconditioned persons may improve CR fitness with only twice-weekly exercise, greater improvement is achieved with a frequency of 3-5 sessions per week."(ACSM 2006)
2. Muscle Strengthening: The dosage requirements for strength training interventions are: a) frequency of 2-3 days per week, b) a minimum of one set of 8-12 repetitions at an intensity of the 8 to 12 Repetition Maximum of each exercise, using any type of strengthening exercise that can be progressed over time.
3. Flexibility Training: Flexibility prescription refers to controlled static stretching in which a subject assumes a position and holds it for a given duration. Dosage requirements are: a) frequency of exercise equal to or greater than two days per week, b) intensity to a position of mild discomfort, c) 3 to 4 repetitions for each stretch held for a duration of 10-30 seconds.
The Physical Activity Guideline of the Centres for Disease Control and Prevention (CDC) (CDC 2001) represents a recommendation supported by epidemiological studies about minimum intensities and duration of physical activity that can improve health-related variables (such as blood pressure and lipid profile). The recommendation that most adults should perform at least 30 minutes of moderate intensity physical activity (in blocks of at least 10 minutes) on five or more days of the week or at least 20 minutes of vigorous intensity exercise at least three days per week, represents a public health statement to the general population. We used the CDC Guideline to evaluate whether interventions had provided an exercise or physical activity stimulus that could improve health.
Two reviewers independently classified studies either as meeting, or not meeting, the ACSM and CDC training guidelines, and then reached a consensus by discussion.
Measures of treatment effect
We planned a priori to consider several variables for outcome measures: pain, tenderness or tender points, global assessments by either patient or physician, the Fibromyalgia Impact Questionnaire (FIQ), anxiety and depression measures, and measures of physical performance and self-efficacy. The outcome measures of interest were most often presented as continuous data with means and standard deviations. Thus, we used a standardized mean difference (SMD), which converts scales to a common metric. To calculate SMD, we used means and standardized deviations of change scores for each intervention. When not available, standard deviations of change scores were derived directly from confidence intervals of change scores, or estimated from the pretest and posttest standard deviations (or standard errors) where these were provided. SMDs were calculated using RevMan Analyses, a statistical analysis tool incorporated in RevMan.
Evaluation of clinically important differences
Recent literature suggests use of relative difference in change scores or percent change as a guide for determining clinically important difference. The Philadelphia Panel, 2001 regards a 15% difference between groups as clinically important and the American College of Rheumatology (ACR) established a definition of improvement in rheumatoid arthritis clinical trials that is 20% in selected measures (Felson 1995). A reduction of approximately 30% in the 11-point pain intensity numerical rating scale (PI-NRS) was found to represent a clinically important difference in clinical trials of chronic pain therapies (Farrar 2001). The PI-NRS study reviewed data from 10 placebo-controlled clinical trials including 2724 subjects, 529 of which had (FM). An additional placebo-controlled study with 99 subjects (Dunkl 2000) found the 11-point PI-NRS, FM Impact Questionnaire (FIQ) and tender point counts were all able to distinguish among groups of patients with FM who reported improved clinical status from those who did not.
Percent change results must be interpreted with caution since estimates of important change have been found to vary by magnitude of baseline measurement on a given scale (Stratford 1998). Disease status or activity level at baseline affects relative improvement when using percent change as a marker for improvement (vanRiel 2000). The scale's sensitivity to change is also dependent on the construct that is used as a comparison for determining clinical importance (Riddle 1998).
We used a conservative estimate of 30% relative percentage improvement as a benchmark for clinical importance based on work by Farrar 2001. Farrar 2001 determined that a reduction of approximately 30% on the pain intensity numerical rating scale represents a clinically important difference for patients with chronic pain (Farrar 2001). This is also consistent with the findings of Dunkl 2000 who examined responsiveness of measures of clinical improvement in FM. Relative percentage improvement was calculated as the mean change in the treatment group minus the mean change in the control group divided by the pooled mean for the baseline scores for the variable.
The Cochrane Musculoskeletal Group has recently adopted new guidelines for describing and interpreting clinical relevance (Tugwell 2004). In this review, in addition to using a criterion of 30%, we have applied the new guidelines. The guidelines were used to develop the plain language summary.
Randomized clinical trials (RCT)
We included RCTs that compared the effects of exercise to the effects of other treatments (e.g., relaxation, Cognitive Behaviour Training) or to control conditions that did not involve any form of activity or treatment (e.g., treatment as usual, attention only, wait list controls). When multiple interventions were compared in a single study, we analysed the comparisons that arose from each exercise intervention separately. In the meta-analysis, we included only the studies which compared exercise to an untreated control.
We preferentially analysed intention to treat (ITT) data whenever available. To create a more complete data set for analysis, we contacted authors of studies with missing data requesting data required for analysis. Several authors generously provided estimates of central tendency or variability (means, standard deviations of pretest and posttest data or standard deviations of change scores) (Buckelew 1998; Burckhardt 1994; DaCosta 2005; Gowans 1999; Richards 2002) and information needed for the qualitative analysis (Altan 2004; Cedraschi 2004; Hakkinen 2001; Mengshoel 1992; Redondo 2004). Studies that only provided categorical data and those for which we were not able to obtain missing data were excluded from meta-analysis (Genc 2002).
Assessment of heterogeneity
We created the data extraction tables and then discussed as a team what issues of clinical heterogeneity should be examined. Important sources of heterogeneity were considered to be: variations in interventions (aerobic training, flexibility training, strength training or mixed/composite training) and dosage of exercise intervention (meeting ASCM training criteria), disparate comparators (e.g., intervention versus a control group or intervention versus a second intervention), and timing of measurement of outcome measures and methodological quality.
Heterogeneity among the trials was next assessed using the heterogeneity statistics (chi squared, I
Assessment of reporting biases
When appropriate, publication bias was assessed using a visual assessment of the funnel plot (RevMan Analyses).
Data synthesis (meta-analysis)
Using RevMan Analyses software, mean change scores were compared and weighted and combined using a random effects model. Where meta-analysis was inappropriate, we used RevMan Analyses to produce effect sizes (SMD) and effect size confidence intervals. We used Cohen's categories for effect size (Cohen 1988) to evaluate the magnitude of the effect (.2 = small effect, .5 = medium effect, .8 = large effect). We used the following levels of evidence descriptors of van Tulder 2003 to classify the results of the meta-analysis:
- Strong - consistent findings among multiple high quality (HQ) RCTs
- Moderate - consistent findings among multiple low quality (LQ) RCTs and/or controlled clinical trials (CCTs) and/or one HQ RCT
- Limited - one LQ RCT and/or CCT
- Conflicting - inconsistent findings among findings among multiple trials (RCTs and/or CCT)
- No evidence from trials - no RCTs or CCTs
We defined inconsistent as:
1. In the absence of high quality studies, at least one RCT clearly favors control while at least one RCT(s) clearly favors treatment
2. If more than one high quality studies is available, and at least one HQ RCT clearly favors control while at least one HQ RCT clearly favors treatment
We defined consistent as:
1. All studies clearly favor treatment
2. All studies clearly favor control
3. Some studies clearly favor treatment, the remainder are inconclusive (do not exclude the null)
We defined "clearly favor" as: the confidence interval excludes zero.
Outcome Measures: When researchers reported more than one measure for a dependent variable, we used the following order of preference for entry into the meta-analysis:
1. Pain: VAS, VAS FIQ, Ordinal Scale
2. Tender Points: dolorimetry, total myalgic score, tender point count
3. Global: FIQ Total, subject-rated VAS or ordinal scale, QOL scale, SIP Total.
4. Depression: Beck Cognitive, Beck Total, CES, FIQ-depression, AIMS Depression
5. Objective Measures of Physical Function: selected on a case-by-case basis depending on the researchers' stated objectives
Subgroup analysis
Subgroup analyses were limited to aerobic-only interventions (from moderate to high quality studies) and strength-only interventions (from poor quality studies). Future updates may include refinements to meta-analysis as new trials come available; for example, we hope to also examine the effects of moderate to high quality strength training and flexibility training as such studies are published.
Sensitivity analysis
Except for the aerobic-only exercise studies, there were too few studies in any other grouping to perform sensitivity analysis. We assessed the bias related to low methodological quality using visual inspection of the forest plots of poor quality studies versus the moderate to high quality aerobic-only studies.
Results
Description of studies
See: Characteristics of included studies; Characteristics of excluded studies.
Results of the search
We inspected 2226 titles generated from the searches conducted in 2002 and 2005 and found 64 citations of full-length articles describing experimental trials which examined the effects of interventions that included an exercise component in subjects with FM.
Included studies
Thirty-four reports (Altan 2004; Buckelew 1998; Burckhardt 1994; Cedraschi 2004; DaCosta 2005; Genc 2002; Meyer 2000; Gowans 1999; Hakkinen 2001; Hakkinen 2002; Isomeri 1993; Jentoft 2001; Jones 2002; Keel 1998; King 2002; Mannerkorpi 2000; Martin 1996; McCain 1988; Mengshoel 1992; Meyer 2000; Nichols 1994; Norregaard 1997; Ramsay 2000; Redondo 2004; Richards 2002; Schachter 2003; Sencan 2004; Valim 2003; Valkeinen 2004; vanSanten 2002a; vanSanten 2002b; Verstappen 1997; Wigers 1996; Zijlstra 2005) met our selection criteria and were included for analysis. Three publications were accompanied by subsequent reports dealing with the same subjects. Hakkinen 2002 reported on additional variables from the Hakkinen 2001 study and thus was counted as one study for analysis. Two publications presented information on long term uncontrolled follow-up of included RCTs: Gowans 2004 was a follow-up to Gowans 2001; Mannerkorpi 2002 was a follow-up to Mannerkorpi 2000; data were not analysed in this review and the reports were treated as secondary studies. The basic characteristics of the included studies are summarized in the 'Characteristics of Included Studies' Table.
Participants
There were a total of 2276 subjects with the confirmed diagnosis of FM across the studies; 1264 subjects were assigned to exercise interventions. The average sample size for the smallest experimental group was 24.7 (SD=16.4, min-max: 5-80) for the 34 original studies. Mean age in the studies ranged from 27.5 to 60.2 years in 34 studies (unspecified in Ramsay 2000). For the 2197 subjects for whom gender was reported, 96.4% were female. Nineteen studies involved only females; twelve examined both males and females, with females in the majority; and the remaining three studies did not specify the gender of participants. Few studies used participation in regular physical activity prior to study entry as an exclusion criterion.
Interventions
The 34 studies in the review comprised 47 interventions that included exercise. Subjects were randomized to at least one aerobic-only intervention in 15 studies, to strength-only interventions in three studies, to flexibility-only interventions in 3 studies, to mixed exercise only intervention in 11 studies, and to a composite aerobic exercise plus education interventions in four studies. Other composite interventions were explored in only one study each: mixed exercise plus medication; flexibility exercise plus medication; mixed exercise plus self management strategies and group discussion; aerobic exercise plus biofeedback; aerobic exercise within a multidisciplinary program; and aerobic exercise as part of a spa treatment. Twelve studies had more than one intervention that included exercise. The details regarding the interventions are provided in Table 2.
Evaluation of Training Stimulus
Twenty studies described exercise interventions that met ACSM recommendations: 17 for aerobic training, three for strength training (Hakkinen 2001; Jentoft 2001; Valkeinen 2004) and two for flexibility (Jones 2002; Mannerkorpi 2000). Eleven of 14 studies that did not meet the ACSM recommendations did not provide sufficient detail about the aerobic, strengthening or flexibility exercises to accurately determine the adequacy of the training stimulus or flexibility intervention. In one study (Norregaard 1997) it was determined that ACSM guidelines for intensity of aerobic training stimulus had not been met because the majority of subjects could not achieve the target heart rate levels (i.e., 40%-50% VO
Only three studies described exercise interventions that met CDC recommendations for aerobic training (McCain 1988; Meyer 2000; Redondo 2004). Seven of 32 interventions that did not meet the CDC recommendations did not provide sufficient detail about the aerobic exercise to accurately determine the adequacy of the training stimulus (Buckelew 1998; Burckhardt 1994; Genc 2002; Isomeri 1993; Keel 1998; Ramsay 2000; Sencan 2004). The remaining 25 studies were insufficient with respect to frequency and/or duration of exercise to satisfy the CDC standards for either moderate (30 min, 5 times/week) or vigorous (20 min, 3 times/week) exercise.
Outcomes
A variety of tests and measures (n=166) were used to evaluate the effects of the six outcomes in the 34 included studies. For example, pain was measured in 28 studies using 9 instruments. Most studies (n = 22) used a 10 cm visual analogue scale to measure pain. The FIQ was the most commonly used test to measure for global well-being (n=13). The most commonly used test to measure physical performance (aerobic) was the 6-minute walk test (n = 6), however seven studies measured maximum oxygen uptake with staged treadmill or cycle ergometer tests. Although several studies used dolorimetry, the most common measure of tender points was the tender point count (12 studies). Depression was measured using the Beck Depression Inventory in five studies and the depression visual analogue in five studies. The most common measure of fatigue was the FIQ fatigue visual analogue scale which was reported in 10 studies.
Adverse Effects
Norregaard 1997 noted that "many patients in the training group actually reported a deterioration of symptoms and did not want to complete the study". Verstappen 1997 commented that, "17% (of completers) reported that their complaints got worse during the intervention period, that the exercise aggravated the feelings of soreness and tiredness afterwards, or that the pain exceeded their tolerance level during the exercises". Mannerkorpi 2000 reported a reduction of planned intensity of exercise because "many patients reported increased pain for 3-4 days after the training sessions". In contrast, Mengshoel 1992 noted that "fibromyalgia patients may perform a low-intensity dynamic endurance exercise ... without exacerbating their general pain and fatigue symptoms". While Richards 2002 reported no adverse effects, they cited increased pain and stiffness as a reason for attrition (for unknown numbers of participants). Schachter 2003 also reported that some participants reported increased pain, stiffness and fatigue (unknown numbers of participants). Wigers 1996 reported that nine subjects experienced increased stress. Aside from these general comments, only five of the 1264 subjects assigned to exercise interventions were designated as having had an adverse effect possibly related to exercise. These included one metatarsal stress fracture (Schachter 2003), one case of ischialgia (Wigers 1996), and two cases of transient knee pain (McCain 1988).
With regards to strength training, Jones 2002 reported a worsening of pain (n=3). Conversely, Hakkinen 2001 did not report any adverse effects and stated "even heavy resistance training can be safely used in the treatment of fibromyalgia" and Valkeinen 2004 reported no adverse effects and commented that "it is noteworthy that, after the initial phase of training, the patients did not complain of any unusual exercise induced pain or muscular soreness during the experimental period, and even intensive strength training did not worsen the symptoms."
Among studies with flexibility exercise interventions, Jones 2002 reported increased pain for some participants while no other researchers reported adverse effects with flexibility training.
In mixed exercise intervention studies, both vanSanten 2002a and vanSanten 2002b describe post exercise soreness as an important barrier to a compliance with mixed exercise training. They noted that despite continuous encouragement, about 50% of the participants in both the high intensity and self-selected intensity mixed exercise training groups were not able to comply with the training sessions and patients in the high intensity group "felt completely 'broken down' for more than 24 hours after the training sessions." Among studies with composite interventions, Cedraschi 2004 speculated that increased pain may have contributed to high attrition rates in the exercise group, but this was not quantified.
Excluded studies
Twenty-seven studies which at first appeared to be appropriate were excluded from this systematic review. Of the 27 studies that were excluded, 4/27 did not adequately characterize the population, 19/27 were not randomized trials, and 4/27 did not include an intervention we would characterize as exercise. (see the 'Characteristics of Excluded Studies' Table).
Studies awaiting assessment
At present there are three new studies are awaiting review: Assis 2006; Gusi 2006; Salek 2005.
Risk of bias in included studies
Methodological Analysis
Ten studies scored 3 out of 5 on the Jadad Scale; 21 scored 2, and two scored 1. The mean of van Tulder quality scores for internal validity was 5.06 (out of a total possible of 11); the mode was 4. The total quality scores for individual studies ranged from 1 to 9. Four studies were classified as high quality ratings, 15 as moderate quality, and 14 as low quality. More than half the studies were deficient in six or more internal validity criteria (concealment of treatment allocation, compliance with treatment, patient blinding, care provider blinding, control of co-intervention, valid randomization). Results of the methodological assessment including the van Tulder item by item analysis and the Jadad scores are provided in Table 3. Table 4 shows the quality assessments for studies that met ACSM dosage guidelines for the aerobic interventions training in the included studies.
Allocation
A method of allocation was specified in 17 studies and treatment allocation was concealed in 10 studies. Allocation methods ranged from computer generated random numbers to simple "by lot" procedures. Treatment allocation was generally concealed by use of an independent researcher who was not aware of the eligibility of the subjects and who had no knowledge of the assignment procedures. There was insufficient information in many studies to determine methods of allocation or any concealment of treatment allocation. With respect to quality of allocation concealment, 10 studies were judged as adequate, there was some uncertainty about adequate concealment in 20 studies and allocation was definitely not concealed in 4 studies.
Blinding
The care provider was blinded to the intervention in only two studies (Richards 2002, Wigers 1996). The subjects were blinded to the intervention in six studies (Buckelew 1998; Keel 1998; McCain 1988; Sencan 2004; vanSanten 2002a; Zijlstra 2005) while the outcome assessors for the main outcomes were blinded in 20 studies. Blinding of the care provider and subjects is difficult during training studies, but blinding of the outcome assessor should be standard practice to reduce the chance of bias in reporting outcomes.
Follow-up and exclusions
Attrition rates for the 17 aerobic exercise intervention groups averaged 27.0% (SD=18.9%, min-max: 0 to 67%); attrition for the two strength training interventions was 0% (SD=0%). In the two flexibility interventions for which attrition data was provided, attrition was 12.5% (SD= 4.9%, min-max: 9 -16%). The 13 mixed exercise interventions had attrition rates of 14.6% (SD=11.8%, min-max: 0-40%). The 11 composite interventions had dropout rates of 14.8% (SD= 9.6%, min-max: 0-27%). Mean attrition in the 20 non-treatment control groups was 12.3% (SD=11.8%, min-max: 0-47%) and in the comparator groups was 18.0% (SD=14.1%, min-max: 0-49%).
Selective reporting
Review authors did not have concerns over selective reporting of time points (i.e., only 2 of the 34 did not have comparability of timing of outcome assessment in across groups). Few studies measured the complete cluster of variables chosen by reviewers for major comparisons. A limited number of studies did not include point estimates and estimates of variability for major outcome measures.
Other potential sources of bias:
Sample size
Small sample size is a methodological weakness of most included studies; only five studies of the 34 included studies meet the standard of 50 subjects per group. In general, there is a trend over time toward larger sample sizes. Mean sample sizes for included studies across 5 year intervals are 18.0, 16.0, 24.0, and 34.9 for 1985-90, 1991-95, 1996-00, 2000-05, respectively.
Adherence
Fundamental to application of exercise as an intervention for FM is the requirement that exercise must be tolerated by individuals with FM. Norregaard 1997 noted "many patients in the training group actually reported a deterioration of symptoms and did not want to complete the study". Verstappen 1997 commented that, "17% (of completers) reported that their complaints got worse during the intervention period, that the exercise aggravated the feelings of soreness and tiredness afterwards, or that the pain exceeded their tolerance level during the exercises". Mannerkorpi 2000 reported a reduction of planned intensity of exercise because "many patients reported increased pain for 3-4 days after the training sessions". Where McCain 1988 reported participants did not have difficulty performing high intensity cycling, vanSanten 2002a reported that 50% of participants both in a similar high intensity program as well as those in a lower intensity program consistent with ACSM guidelines were unable to fully comply with the training sessions and that participant in the high intensity group reported feeling "broken-down". Such difficulties with exercise combined with high attrition rates in exercise studies suggest the importance of identifying the dose-response curves for FM signs and symptoms for different types (and to a lesser extent, modes) of exercise. One important tool used in the examination of these questions is exercise adherence. To accurately describe the dose-response relationship for exercise training, detailed data on the frequency, intensity and duration of performed exercise are required.
A common tool used to represent exercise adherence in included studies is the reporting of exercise frequency, as measured by total attendance at sessions (in the case of organized exercise sessions) or participant-reported performance of exercise sessions (in the case of independent exercise programs such as home exercise programs). Focusing on studies that used aerobic training or included an aerobic training component within a mixed exercise or composite intervention, eighteen studies monitored attendance at organized exercise sessions or performance of independent exercise sessions. In using frequency as a measure of exercise adherence, authors operate under the assumption that participants performed all of the prescribed exercise (duration) at the prescribed intensity. Yet difficulties cited in the preceding paragraph suggest that this assumption may be inaccurate, that the intensity and duration of performed exercise may differ greatly from the prescribed exercise.
Taking further steps to identify the dose-response curves for exercise requires employment of techniques to monitor exercise duration and intensity. Focusing on studies that used aerobic training or incorporated an aerobic training component within a mixed or composite intervention, fourteen measured intensity using rating of perceived exertion or heart rate (via measurement of pulse by participant, heart rate or pulse monitors). No studies reported whether participants completed the prescribed duration of exercise in each session. No studies reported detailed results of systematic data collection and analysis of participant adherence to exercise performance in a way that would allow readers to understand the exact training volume achieved by participants.
In studies of strength training, adherence can be assessed through a combination of attendance/performance of independent sessions and the monitoring of sets and repetitions of exercises. Focusing on studies that evaluated strength training, one study reported on attendance at organized exercise sessions. No studies reported on performed sets and repetitions of exercises.
Effects of interventions
The results will be presented beginning with the meta-analysis, followed by the analyses of standardized mean differences (effect sizes), and description of relative percentage change (clinically significant difference).
1. Meta-analysis
After preliminary meta-analyses in which we examined sources of statistical heterogeneity, clinical heterogeneity and methodological quality, we meta-analyzed the data for effects of exercise on pain, global wellbeing, and objective measures of physical function for aerobic-only interventions and strength-only interventions when compared to untreated control groups. We further restricted our meta-analysis to moderate and high quality studies for evaluation of treatment effects of aerobic-only interventions. The results are displayed with significance level, standardized mean difference (SMD) with confidence intervals and forest plots.
Meta-analyses of Aerobic-Only Exercise Interventions compared to Untreated Control Groups
There is moderate evidence that short term (6 to 23 weeks) aerobic-only exercise training prescribed at ACSM levels results in:
- Moderately large positive (but statistically insignificant) effects on pain: SMD = 0.65 (95%CI: -0.09 to 1.39) pooled from 183 subjects in one high quality (Wigers 1996) and two moderate quality studies (Buckelew 1998; Schachter 2003)
- Medium-size positive effects on global well-being: SMD = 0.49 (95%CI: 0.23 to 0.75) pooled from 269 subjects from four moderate quality studies (Buckelew 1998; Gowans 2001; King 2002; Schachter 2003)
- Moderately large positive effects on objective measures of physical function: SMD = 0.66 (95%CI: 0.41 to 0.92) pooled from 253 subjects in one high quality (Wigers 1996) and three moderate quality studies (Gowans 2001; King 2002; Schachter 2003)
- Small, statistically insignificant effect for tender points: Although the pooling of data from 309 subjects is non-significant: SMD = 0.23 (95%CI: -0.18 to 0.65), one high quality, Wigers 1996 reported significant large effects, while three moderate quality studies (Buckelew 1998;Gowans 2001;Schachter 2003) found a small non-significant positive effect and another moderate quality study (King 2002) found a small non-significant negative effect.
There is conflicting evidence that the effect of short-term (6 to 23 weeks) aerobic-only exercise training prescribed at ACSM levels results in:
- Small to medium improvement in depression: SMD = 0.40 (95%CI: 0.04 to 0.76) pooled from 233 subjects from one high quality (Wigers 1996) and one medium quality study (Schachter 2003) which showed no evidence of effect, and two moderate quality studies (Buckelew 1998; Gowans 2001) which demonstrated medium to large effects.
There is limited evidence (one medium quality study, Schachter 2003, 87 subjects) that 16 weeks of Aerobic Only prescribed at ACSM levels has no effect in individuals with FM on stiffness: SMD = -0.17 (95%CI: -0.59 to 0.25) or fatigue: SMD = 0.00 (95%CI: -0.52 to 0.52).
There is limited evidence from one moderate quality study (Sencan 2004, 40 subjects) that short term (6 weeks) aerobic-only exercise training prescribed at an unspecified intensity compared to an untreated control produces large improvements in pain: SMD = 1.34 (95% CI: 0.65 to 2.04), tender points: SMD = 3.90 (95% CI: 2.80 to 4.99), and depression: SMD = 1.22 (95% CI: 0.54 to 1.90).
Meta-analyses of Strength-Only Exercise Interventions compared to Untreated Control Groups
There is limited evidence from two low quality studies (Hakkinen 2001; Valkeinen 2004) that 21 weeks of Strengthening Exercise Only prescribed at ACSM levels versus Untreated Control for individuals with FM results in:
- a large effect on pain (Hakkinen 2001, 21 subjects, SMD = 3.00, 95%CI: 1.68 to 4.32)
- a large effect on global wellbeing (Hakkinen 2001;Valkeinen 2004, 47 subjects, SMD = 1.43, 95%CI: 0.76 to 2.10)
- a medium (non-significant) effect on objective measures of physical function (Hakkinen 2001; Valkeinen 2004, 47 subjects, SMD = 0.52, 95%CI: -0.07 to 1.10).
- a large effect on tender points (Valkeinen 2004, 26 subjects, SMD = 1.52, 95%CI: 0.63 to 2.41)
- a large effect on depression (Hakkinen 2001, 21 subjects, SMD = 1.14, 95%CI: 0.20 to 2.08)
- no differences in magnitude and time course of adaptations of the neuromuscular system to strength training in 11 women with FM (Hakkinen 2002) as compared to 10 healthy women. The authors also found a similar response in systematic in growth hormone levels during an acute bout of exercise in participants with FM and healthy controls.
2. Effectiveness of all other interventions in included studies
Several interventions were excluded from the meta-analyses because of inadequate replication or clinical heterogeneity. The standardized mean differences (effect sizes) and confidence intervals for the effects of these exercise-only interventions and the composite interventions on global well-being, pain, tender points, observer-measured physical function and depression are summarized in Table 5 and Table 6. Several statistically significant medium to large effect sizes (equal to or greater than .5 with 95% confidence interval excluding zero) were identified favouring several exercise-only and composite interventions. In terms of frequency, global well-being was the most commonly identified as having a statistically significant medium to large effect size; six interventions produced significant effect sizes of this magnitude for variables addressing global change.
3. Relative Percentage Change
The calculation of relative percentage change revealed several sporadic clinically significant improvements in a variety of outcomes.
- Aerobics Only Interventions: Clinically significant improvement (>30% greater change at post-test) in five aerobics-only interventions as compared to an untreated control in depression (Buckelew 1998, Wigers 1996), tender points (Sencan 2004, Wigers 1996), in FIQ Total (Schachter 2003), dynamic endurance (Mengshoel 1992), FIQ rested and self-efficacy for function (Schachter 2003).
- Strength Only Interventions: Greater than 30% improvement was seen in the strength training group as compared to an untreated control group in pain (Hakkinen 2001), global well-being (Hakkinen 2001, Valkeinen 2004), and depression (Hakkinen 2001).
- Other Interventions: Mannerkorpi 2000 found clinically significant improvement in anxiety, SF-36 general health, SF-36 role physical, and SF-36 vitality. Clinically significant improvement was observed in anxiety and depression from composite program of aerobics exercise and education when compared to an untreated control group in two studies (Cedraschi 2004, Zijlstra 2005). Altan 2004 and Buckelew 1998 found clinically significant improvement in depression resulting from mixed exercise and composite aerobics and biofeedback (respectively) compared to an untreated control group. Cedraschi 2004 and Zijlstra 2005 found clinically significant improvement in FIQ worked missed resulting from composite interventions of aerobic and spa and aerobics and education as compared to untreated control groups.
In addition to these findings, the results of the detailed analysis of clinical relevance using the 2006 Cochrane Musculoskeletal Group guidelines are presented in Table 7 and Table 8.
4. Long-term effects
Most studies examined the effects of short-term interventions - interventions varied between 2.5 weeks to 24 weeks (quartiles of 8, 12, and 20 weeks). Seven aerobic only studies reported follow-up assessment varying from 12 weeks to 4 years. At 12 weeks, King 2002 found no between group differences between the aerobics only group and the control. At 24 weeks post intervention, improvements in pain pressure threshold, depression, and pain were maintained (Sencan 2004). Richards 2002 reported improved FIQ scores at 24 weeks, and fewer active tender points in the aerobic group compared to the relaxation group at 1 year. At one year follow-up, DaCosta 2005 reported improved upper body pain and FIQ scores. Gowans 1999 conducted a program review of participants 3 to 6 months after completion of the intervention (no control group was used in the follow up) and reported significant improvements in 6 minute walk, fatigue, self-efficacy for pain and symptoms. Gowans 2004 noted in an uncontrolled follow-up of Gowans 2001, that improvements in physical function (6 min walk test) mood (Beck Depression Inventory - total, Beck Depression Inventory - somatic), self-efficacy (function), and symptom severity (FIQ total) were maintained at 6 and 12 month (ITT analysis). Buckelew 1998 included a two year maintenance phase in which interventions were monitored on a monthly basis and exercise was performed as a home program, and which lost very few participants to follow up. Improvements in self-reported physical function and self-efficacy for function were seen at one year follow up. Wigers 1996 evaluated participants 4.5 years after the intervention and found that improvements were not retained in the exercise group, and that most of the group was not still exercising.
In an uncontrolled follow-up to Mannerkorpi 2000, Mannerkorpi 2002 noted that participants in the exercise group had maintained improvements in FIQ total, FIQ physical function, SF-36 physical function, SF-36 general health, and grip strength to six months, and FIQ pain, FIQ fatigue, physical function (6 min walk test) and SF-36 bodily pain, social function, vitality to 24 months. Zijlstra 2005 noted that at 3 months follow-up, the spa group were significantly better than controls in the physical component of health status measure RAND-36, pain, fatigue, subject evaluated general health, number of tender points, graded tender point score, and FIQ total. Cedraschi 2004 noted that at 24 months follow-up, significantly more patients in the treatment group than in the control group had engaged in a new physical activity. At 12 week follow-up, Altan 2004 found improvements in depression inventory were retained; there were no between group differences in symptoms (FIQ, pain, fatigue, stiffness, sleep), tender points, patient- or physician-rated global health status, or in chair test either immediately following the intervention or at the 12 week follow-up. Neither Jentoft 2001 nor Redondo 2004 found any significant between-group differences at 46 and 52 weeks respectively. Ramsay 2000 who had only one significant between group difference at the end of intervention (HAD-anxiety) between a weekly supervised aerobic exercise class and an unsupervised home program found no between group differences at 24 or 48 weeks.
5. Adverse Effects
Norregaard 1997 noted that "many patients in the training group actually reported a deterioration of symptoms and did not want to complete the study". Verstappen 1997 commented that, "17% (of completers) reported that their complaints got worse during the intervention period, that the exercise aggravated the feelings of soreness and tiredness afterwards, or that the pain exceeded their tolerance level during the exercises". Mannerkorpi 2000 reported a reduction of planned intensity of exercise because "many patients reported increased pain for 3-4 days after the training sessions". In contrast, Mengshoel 1992 noted that "fibromyalgia patients may perform a low-intensity dynamic endurance exercise ... without exacerbating their general pain and fatigue symptoms". While Richards 2002 reported no adverse effects, they cited increased pain and stiffness as a reason for attrition (for unknown numbers of participants). Schachter 2003 also reported that some participants reported increased pain, stiffness and fatigue (unknown numbers of participants). Wigers 1996 reported that nine subjects experienced increased stress. Aside from these general comments, only five of the 1264 subjects assigned to exercise interventions were designated as having had an adverse effect possibly related to exercise. These included one metatarsal stress fracture (Schachter 2003), one case of ischialgia (Wigers 1996), and two cases of transient knee pain (McCain 1988).
With regards to strength training, Jones 2002 reported a worsening of pain (n=3). Conversely, Hakkinen 2001 did not report any adverse effects and stated "even heavy resistance training can be safely used in the treatment of fibromyalgia". Valkeinen 2004 reported no adverse effects and commented that "it is noteworthy that, after the initial phase of training, the patients did not complain of any unusual exercise induced pain or muscular soreness during the experimental period, and even intensive strength training did not worsen the symptoms."
Among studies with flexibility exercise interventions, Jones 2002 reported increased pain for some participants while no other researchers reported adverse effects with flexibility training.
In mixed exercise intervention studies, both vanSanten 2002a and vanSanten 2002b describe post exercise soreness as an important barrier to a compliance with mixed exercise training. They noted that despite continuous encouragement, about 50% of the participants in both the high intensity and self-selected intensity mixed exercise training groups were not able to comply with the training sessions and patients in the high intensity group "felt completely 'broken down' for more than 24 hours after the training sessions." Among studies with composite interventions, Cedraschi 2004 speculated that increased pain may have contributed to high attrition rates in the exercise group, but this was not quantified.
Discussion
Summary of main results (benefits and harms)
The main results of our review are as follows:
1) Moderate quality evidence exists that aerobic-only exercise training at recommended intensity levels has medium-sized positive effects on global outcome measures and medium-sized positive effects on measures of physical function; the effect of such exercise on pain and tender points is less certain but we cannot rule out the possibility that aerobic exercise has a large positive effect on these variables;
2) Strength and flexibility exercise remain under-evaluated as exercise prescriptions for people with FM;
3) Despite the increasing number of studies investigating the effect of combination treatments with exercise, this question has also received inadequate study, principally because there is underutilization of appropriate research designs;
4) There is limited evidence on a variety of other outcomes including stiffness, fatigue, and depression;
5) The analysis of percentage change reinforced the results of meta-analysis.
In aerobics-only interventions, clinically significant improvements were found sporadically in six variables: depression, tender points, global well-being, physical function, self-efficacy and symptoms.
Aerobic-only
The evidence has been steadily accumulating to the degree that we were now able to meta-analyze a body of six studies of moderate to high quality comparing aerobic-only protocols of ASCM-recommended intensity to an untreated control group. With respect to our a priori primary outcomes, we identified the existence of moderate quality evidence that short-term aerobic-only exercise training at recommended intensity levels (ACSM 2006) produced medium-sized positive effects on global outcome measures (Buckelew 1998, Gowans 2001, King 2002, Schachter 2003 ) and medium-sized positive effects on measures of physical function (Gowans 2001, King 2002, Schachter 2003, Wigers 1996). Effects on pain and tender points were statistically insignificant. With respect to the secondary outcomes (depression, fatigue/sleep), there is conflicting evidence regarding the effect of aerobic exercise on depression, with two studies reporting medium to large effects (Buckelew 1998, Gowans 2001) and two others, no evidence of effect (Schachter 2003, Wigers 1996). There is limited evidence from one medium quality study (Schachter 2003) that aerobic-only exercise prescribed at ACSM levels has no effect on fatigue in individuals with FM.
Although the meta-analysis indicates there is a positive effect of aerobic exercise on global well-being and physical function, our appraisal is moderated by several factors. Attrition rates were high in most studies (range 13-44%) and all studies had relatively small sample sizes (range 16-51). In addition, although adherence is rarely well documented, there many indications that adherence to both exercise intensity and frequency is poor. It must also be noted that more high-quality studies are needed as our analysis only includes one such study (Wigers 1996) and that study had a small sample size (n=16 in the intervention group).
Strength-only
Currently, there is limited evidence that strength-only exercise has a large effect on pain (Hakkinen 2001), global well-being (Hakkinen 2001, Valkeinen 2004), physical function (Hakkinen 2001, Valkeinen 2004), tender points (Valkeinen 2004) and depression (Hakkinen 2001). There is also limited evidence that strength-only exercise compared to flexibility exercise has a medium size positive effect on pain and global well-being, but no effect on muscle strength (Jones 2002). However, only one of these studies was even of medium quality (Jones 2002) and all three had small sample sizes (range 11-28) and inadequate adherence reporting information. Of interest, Hakkinen 2001and Valkeinen 2004 reported the strength-only exercise had no adverse effects and zero attrition, which is a promising finding given the mean attrition rate of 23.3% reported in the aerobic-only studies. Despite some intriguing results from the available literature however, at this time we are unable to recommend strength training past the suggestion that additional studies, preferably of high quality, are needed to confirm the effect of such exercise on people with FM.
Flexibility-only
There were no studies comparing flexibility-only exercise to a non-exercise control group. There is limited evidence from one moderate quality study (Jones 2002) that flexibility-only exercise compared to strength exercise has a large positive effect on flexibility, but no effect on tender points or depression. Thus more studies, again preferably of high quality, are needed to confirm and extend the effects of flexibility exercise.
Mixed Studies
There were not enough studies to enable meta-analysis of mixed-exercise studies, in which comparisons included untreated controls, high versus low intensity, water versus land, relaxation, balneotherapy (immersion of part or all of the body in a mineral water bath), and Cognitive Behaviour Training. Without exception, the evidence for any particular comparison was limited and arose from a single medium quality study. There were no positive effects of any mixed intervention on signs and symptoms of FM with the exception of objective measures of physical function (vanSanten 2002a, versus untreated control; Martin 1996, versus relaxation).
Composite studies
Composite studies were widely varied, with interventions including Aerobic-and-Education, Exercise-and-Self Help, Mixed Exercise-and-Education, Exercise-and-Spa Treatment, and Exercise-and-Self Management Strategies. Comparisons included untreated controls, education and relaxation. Evidence for effects was limited for every comparison, with a mix of positive and no-difference effects for the active intervention. Only in the case of Aerobics-and-Education versus untreated control did more than one study (Burckhardt 1994, Gowans 1999; both low quality) contribute to the results that indicated limited evidence of no difference in pain and a medium effect on measures of physical function in favour of the intervention.
A concern with all the composite studies is that the experimental designs did not permit a comparison between the individual composite interventions and control groups. Thus there is no way to determine the independent effect of, for example, aerobic exercise or education, or their potential interaction. Use of 2x2 factorial designs is recommended for all such composite studies to maximize information. While this would add to the complexity of organizing the study and require more participants, such designs would help to clarify the effects of both composite and "pure" interventions.
Relative Change
Our analysis across all interventions revealed only sporadic occurrences of clinically significant improvements (defined as > 30%, Farrar 2001), and these improvements were observed only when exercise interventions were compared to untreated control groups. In reference to our primary outcome variables, clinically significant improvements were observed and reported in: a) pain in one strength-only intervention; b) global well-being in one aerobic-only, two strength-only, and one composite intervention; c) physical performance-aerobic in no interventions, physical performance-musculoskeletal in one aerobics-only intervention, and physical performance-flexibility in no interventions; and d) in tender points in two aerobic-only interventions. Clinically significant improvement was found in the following secondary outcome measures: a) depression in seven studies (two aerobic, one strength-only, one mixed exercise-only, and three composite interventions); b) fatigue in no interventions; and c) sleep (rested in morning) in one aerobic-only intervention. A 30% relative improvement probably aligns with a large effect size. These sporadic clinically significant effects were in general agreement with the occurrences of moderate to large effects in the meta-analyses and the effect size analyses.
Harms
Adverse effects of exercise were not always well reported in studies. Available information was quite variable, with some studies reporting no adverse effects, some reporting vague effects (e.g. increased stress) and others attributing high attrition rates to exacerbations of typical signs and symptoms of FM. On occasion it seemed to be difficult to separate adverse effects caused by the exercise protocol from normal fluctuations in signs and symptoms of FM (e.g. Richards 2002). The reported adverse effects suggest that researchers regarded adverse effects primarily as problems outside of the sphere of symptoms of fibromyalgia. It is possible that participants experiencing increased FM symptoms regarded them as adverse effects to exercise. Verstappen 1997 for example reported that 17% of completers/people who responded to a questionnaire about perceived benefits of the exercise program felt "that their complaints got worse during the intervention period, that the exercise aggravated the feelings of soreness and tiredness afterwards, or that the pain exceeded the tolerance level" (pp. 23. Greater attention to ensuring that a definition of adverse effects is shared by researchers and participants is warranted in the future.
Overall, it is our impression that most exercise programs reviewed in this update can be safely completed by people with FM. Clinicians prescribing exercise are reminded to increase intensity slowly, check frequently with participants for adverse effects, and be prepared to back off the exercise program until such effects subside.
Overall completeness and applicability of evidence
There is still a lack of studies, particularly good quality studies, to fully meet the objectives of this review. With the exception of aerobic-only exercise, there are only one or two studies at any level of quality that address many of our specific comparisons. This may be in part because of the large number of exercise modes that have been employed and also in part because research has not been guided by convincing results in previous literature. Fibromyalgia syndrome is also a difficult syndrome to study because of the exacerbations and accompanying signs and symptoms that make participation in exercise problematic for many people. However, with respect to aerobic-only exercise, we now have a pool of six moderate to high quality studies that do permit meta-analysis, with results indicating a positive effect of such exercise on important outcomes.
With respect to participants, studies continue to overwhelmingly recruit women, as per the well-established demographics of FM, but studies of men with FM would provide welcome information. There are overly numerous interventions; the issues around mixed and composite interventions have been identified above. There is also a bewildering array of outcome measures assessing the same constructs that may contribute to the large variability in some of our meta-analyses.
The primary objective of this systematic review was to evaluate the effects of exercise training including cardiorespiratory (aerobic), muscle strengthening, and/or flexibility exercise on global well-being, selected signs and symptoms, and physical function in individuals with FM. We have critically appraised and summarized 34 studies, all of which are relevant to this question, at least to some degree. Despite the noted limitations in study quality and sample size, we are thus confident of the external validity of our review, that is to say, this review is a valid reflection of the current literature on the effects of exercise on FM.
Methodological Quality
Quality of the evidence
We reviewed 21 new studies in this update, for a cumulative total of 34 including the first review (Busch 2002). Despite this larger pool, there are still too few studies to allow definitive conclusions regarding the effect of strength or flexibility exercise in people with FM. However there are now six studies, of moderate to high quality that have investigated the effect of short-term aerobic-only exercise (at recommended intensity levels) in people with FM. These studies permit the conclusion (derived from meta-analysis) that such exercise has medium-size positive effects on pain, small to medium-sized positive effects on global outcome measures and medium-sized positive effects on measures of physical function. The sample size in these six studies ranged from 16 to 51 in the intervention groups. Their internal validity scores on the van Tulder scale ranged from 5-9, and they scored between 2 and 3 on the Jadad scale. Attrition ranged between 13 and 44% and adherence to the prescribed training programs was generally poorly reported.
Adherence
The studies reviewed do not present a sufficiently clear picture of: a) the intensity and progression of exercise prescribed by researchers, or b) the intensity and progression of exercise performed by the participants. Intended target exercise intensities were expressed in broad terms (for example, 60-80% of age predicted maximal heart rate, as in Martin 1996). This represents a large range of intensities: 60% of maximum heart rate is now considered light intensity exercise, while 80% is considered to be within the low end of the range classified as hard intensity (ACSM 2006). Greater detail about exercise intensity (such as presented in Meyer 2000) is desirable. In addition, consistent use of one system for the classification of exercise intensity is needed in future studies so that researchers, clinicians and individuals with FM have the same understanding of light, moderate and vigorous intensity exercise. Furthermore, the studies reviewed do not present a consistent picture of actual level of exercise performed. No study analyzed and reported intensity of exercise performed in a systematic manner. Without further reporting of adherence to targets within each exercise session, the reviewers were left without a definitive understanding of intensities of exercise tolerated by individuals with fibromyalgia. Researchers such as McCain 1988 presented a picture in which individuals with fibromyalgia can successfully perform vigorous aerobic training. In contrast, a number of researchers noted poor adherence to the prescribed exercise. Meyer 2000 used exercise logs to monitor adherence to individual sessions and stated that "it appears that whereas exercising at high intensity is tolerated for a limited duration over a short period of time, prolonged exposure to a high-intensity exercise regime is not well tolerated and may have an adverse effect on function. On the other hand, exercising at a low-intensity level is well tolerated and may be associated with improved function as adherence continues." vanSanten 2002a supported Meyer's findings but also reported that 50% of participants following a less vigorous exercise prescription had difficulty performing the prescribed exercise intensity. In Norregaard 1997, the researchers stated that while the planned level of performance of the aerobic dance program corresponded to 40-50% VO
Systematic monitoring and reporting of exercise intensity, duration, and frequency for each mode of exercise examined, is important to more clearly identify the types and modes of exercise best tolerated by individuals with fibromyalgia as well as to determine any dose-response curve for various exercise types and to a lesser extent, within each type, modes of exercise. Examining participants' impressions of, and reactions to exercise, with special attention to those who drop out of programs, may yield valuable information that would help clinicians understand more fully the impact of, and problems with, exercise for individuals with FM. More detailed reporting of the exercise program and adherence to exercise sessions is required to explore this more fully in order to design future programs with improved adherence and lower attrition rates among participants.
The intensity of aerobic exercise in the six studies in the aerobic-only meta-analysis does comply with the ACSM guidelines but generally falls into the low domain of the recommended intensity range. Progression of intensity and/or duration for the treatment protocols was largely unsuccessful. Thus, these studies involved intensity levels that fall into that described by ACSM guidelines as appropriate for the early part of an aerobic training protocol (i.e. the first 4 weeks). Typically, greater initial improvement is expected when sedentary, deconditioned individuals become more active; beyond that, unless the exercise regimen is progressed, a tapering off of improvements would be expected. Because training volume did not increase in many of the included studies, it is unclear if the initial reported improvements would continue if programs had continued for longer duration.
Heterogeneity
The effect sizes for our primary outcomes were small to medium and several I
Visual scanning of forest plots for subgroup analysis suggested a random distribution with results in the same direction for most subgroups of studies, suggesting that though the study effects differed in size, their results were mostly in the same direction. This is supported by the P-value tests of significance associated with the meta-analysis.
Sample size
There is no agreement as to the sample size needed to demonstrate clinically important effects. In this review, most of the studies were underpowered and few formal power calculations were reported. None of the intervention groups was larger than 85 subjects per group, the median of the size of intervention groups 22.5 (min= 5, Q1=15.75, Q3=30.5, max = 84). Therefore, we must be cautious when making decisions about "no effect". The average sample size for the smallest experimental group was 24.7 (SD=16.4, range: 5-80) for the 34 original studies. Similar concerns about sample size were raised in our initial review (Busch 2002) and it is somewhat disheartening to see that little progress has been made in the intervening period. However, mean sample size for the intervention groups in our initial review (16 studies) was 19.4 subjects; in the additional 18 studies since 2002, mean sample size increased to 39, so this must be considered a positive trend towards reaching more fully-powered studies.
Outcome measures
There has also been little progress with respect to developing any agreement around a common set of tests and measures for this population. In our initial review (16 studies), more than 60 instruments were used to evaluate outcomes (Busch 2002). In this update (19 studies) more than 100 tests and instruments were used including eight for pain, nine for cardiovascular fitness and thirteen for physical function. The plethora of tests and measures to evaluate outcomes make it difficult to combine data via meta-analysis without inducing heterogeneity, although many of the measures have acceptable levels of reliability and validity. Nevertheless, it would improve consistency of reporting and confidence in analysis if there were more agreement around "gold standards" for assessing outcomes of interest. We thus repeat our previous call for establishing a core set of outcome measures for research into non-medical treatments for FM.
Potential biases in the review process
There are limitations inherent in the primary literature including incomplete description of the exercise protocols, inadequate sample sizes, inappropriate designs for assessing composite exercise programs, and inadequate documentation of adherence to exercise prescriptions.
In our review process, we attempted to control for biases as follows:
- we did not limit our search to English-only publications
- we assessed the reliability of our selection of relevant studies and our grading of the quality of the studies
- we contacted primary authors for clarification and additional information where indicated, although responses were not always obtained
- we examined clinical sources of heterogeneity
- our description of the results was based on a careful consideration of intervention characteristics, study population, methodologic rigour, pre-identification of levels of evidence and group discussion of evidence tables to reach consensus
- we used a multi-disciplinary team with expertise in critical appraisal, pain, clinical rheumatology, physical therapy, exercise physiology and knowledge translation
- where researchers evaluated treatment effects at multiple points, we used the data points closest to 12 weeks to standardize our comparisons.
Agreements and disagreements with other studies or reviews
Our findings are supported in the work of other reviews. Goldenberg 2004 examined overall management of FM and concluded that among studies comparing exercise to wait-list or flexibility controls but not blinded trials, there is strong evidence for efficacy for the use of cardiovascular exercise as part of symptom management and moderate evidence for efficacy of strength training. The authors go on to say, "a stepwise program emphasizing education, certain medications, exercise, cognitive therapy, or all 4 should be recommended" (P 2388). Differences between Goldenberg 2004 and our review can be explained by the classification of evidence (Goldenberg 2004 has classified strong evidence as positive results from a meta-analysis or consistently positive results from more than one RCT and moderate evidence as positive results from 1 RCT or consistently positive results from multiple non-RCT studies), and from the Goldenberg's inclusion of flexibility training as a control treatment.
In the Guideline for the Management of Fibromyalgia Syndrome Pain in Adults and Children (Burckhardt 2005), the American Pain Society also recommend multifaceted management that combines pharmacologic and non-pharmacologic therapies. The latter includes cognitive-behavior therapy that includes self-management strategies, aerobic exercise and clinician-assisted treatments such as hypnosis, acupuncture, therapeutic message and chiropractic manipulation. The American Pain Society Guidelines 'strongly encourage' moderately intense aerobic exercise at least two or three times a week using a slow and gradual progression to avoid exercise-induced pain exacerbation and possible discontinuation of the exercise. The American Pain Society Guidelines are consistent with findings of this review.
In their systematic review, Mannerkorpi 2003 report that low intensity aerobic exercise, such as walking, can improve function and symptoms, that moderate intensity exercise can improve aerobic capacity and reduce tenderness but that high intensity exercise be undertaken with caution. They also report that strength training can improve strength without exacerbation of symptoms. These findings are consistent with the current review. The absence of information about the search strategy used by these authors prohibits further comparisons with findings of this study.
In a systematic review of mind body therapies (MBT) for FM that included interventions such as cognitive behavioral therapy, biofeedback and relaxation training, Hadhazy 2000 found exercise to be superior to MBT for pain and function. Examining interventions that included MBT plus exercise versus controls, the reviewers found moderate evidence of improvement in self-efficacy and quality of life favoring the combined intervention and limited evidence for the intervention for all other outcomes.
Authors' conclusions
Implications for practice There is moderate quality evidence that short-term aerobic training (at the intensity recommended for increases in cardiorespiratory fitness) produces important benefits in people with FM in global outcome measures, physical function, and possibly pain and tender points There is limited evidence that strength training improves a number of outcomes including pain, global wellbeing, physical function, tender points and depression. There is insufficient evidence regarding the effects of flexibility exercise. Adherence to many of the aerobic exercise interventions described in the included studies was poor. |
Implications for research Numerous implications for further research arose from this review. They include:
|
Acknowledgements
We would like to acknowledge the following:
Louise Falzon for help with the literature search, Lara Maxwell and Nancy Santesso of the Cochrane Musculoskeletal Review Group for help with numerous procedural issues and questions, Laurel Walker, Jessica Peterson, Chantale Nightingale, Stephanie Hogan for help with manuscript work, and Claire Bombardier, co-author on the previous Cochrane review.
Data and analyses
- Top of page
- Background
- Objectives
- Methods
- Results
- Discussion
- Authors' conclusions
- Acknowledgements
- Data and analyses
- What's new
- History
- Contributions of authors
- Declarations of interest
- Sources of support
- Notes
- Index terms
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||
What's new
Last assessed as up-to-date: 16 August 2007.
| |||||||||
History
Review first published: Issue 2, 2002
| |||||||||
Contributions of authors
AJB: designing and reviewing protocol for review, performing literature search, data extraction, methodological analysis, writing and reviewing manuscript
CLS: designing and reviewing protocol for review, performing literature search, data extraction, methodological analysis, writing and reviewing manuscript
PMP: designing and writing protocol for review, performing literature search, methodological analysis, statistical analysis, writing and reviewing manuscript
KARB: data extraction, methodological analysis, writing and reviewing manuscript
TJO: data extraction, methodological analysis, writing and reviewing manuscript
Declarations of interest
We confirm that any present or past affiliations or other involvement in any organisation or entity with an interest in the Review which might lead me/us to have a real or perceived conflict of interest are listed below.
- None known
Sources of support
Internal sources
- School of Physical Therapy, University of Saskatchewan, Canada.
- Department of Medicine, University of Saskatchewan, Canada.
- Institute of Health and Outcomes Research, University of Saskatchewan, Canada.
- Institute for Work and Health, Canada.
External sources
- No sources of support supplied
Notes
This review is a major update of the previous review completed in 2002. Methodological differences between the previous review and this update included small revisions to the search terms and changes in the membership of the review team. Also in this update, 11 items of the van Tulder (2003) methodological criteria that reflect internal validity were used to classify studies into high, moderate and low quality studies. In the data synthesis, greater weight was placed on moderate to high quality studies comparing exercise-only interventions to control groups.
The previous review was based on 16 randomized trials whereas this was based on 34 randomized trials. To aid in the interpretation of a growing number of interventions and comparisons, in this update we have expanded the analysis to include calculation of relative percentage change and standardized mean differences. Despite the increased number of reports, meta-analysis continued to be restricted due to clinical heterogeneity and conclusions have not changed substantially.
Index terms
Medical Subject Headings (MeSH)
*Exercise; Exercise Tolerance; Fibromyalgia [*rehabilitation]; Randomized Controlled Trials as Topic
MeSH check words
Humans
* Indicates the major publication for the study