Results of the review are described and commented upon in Table 1 (see Additional file 1). Studies are listed in chronological order by year. The following paragraphs summarize the findings by each column in the table.

Through December 2005, 3035 subjects participated in an FM exercise study. Of that number, 2888 (2400 women, 73 men, 415 gender not reported) were patients with FM. Control subjects with various chronic diseases other than FM numbered 135 and there were 12 healthy control subjects. The ages of subjects ranged from 18–80 years, with a mean of 49.5 years. Older adults, men, and minority persons were underrepresented and no exercise interventions with children were found.

Most interventions were comprised of the three major modes of exercise (aerobic training, strength training, flexibility) either singly or in combination. Aerobic training included cycling, walking, calisthenics, pool exercise or dance. Thirty-three were land-based. Seven were exclusively water/pool based while the remainder either progressed from water to land based or mixed water and land throughout the intervention. Five studies used only weight training, either machine weights or free weights (hand weights/elastic bands). Progression was determined by changes in number of repetitions (reps), sets or increasing load (e.g., progressive 8–20 reps, 4–6 sets, and load increased by 40–80%) or progression of elastic band tension. Three studies tested flexibility as either the active intervention or the control intervention. Stretches were described as static and progressed by self-limited tension and discomfort or an increase in time from 10–90 seconds per major muscle group. Three studies tested the independent and combined effect of a drug and exercise (amitriptyline and pyridostigmine). Four used movement therapies (e.g. T'ai Chi, QiGong, balneotherapy, thalassotherapy). We acknowledge that other therapies using balneotherapy in FM exist as a modality for treating symptoms, but are not included in this review as they were not combined with exercise.

Aerobic intensity was reported in 14 studies as target heart rate or percent age-predicted maximum heart rate determined by standard equations. No study set work rate based on initial maximal aerobic capacity determined by graded exercise test. Sustained target heart rate goals ranged from 120–150 beats per minute. Percent maximum heart rates were usually progressive and ranged from 40%–80% of age-predicted maximum. Borg's Rating of Perceived Exertion (RPE) scale or the "ability to talk test" was used in two studies. Measuring heart rate was most often accomplished by self-assessed pulse rate or less frequently by heart rate telemetry.

The number of exercise sessions ranged from 1–5 times per week most commonly 2–3 times weekly. Length of class time ranged from 15–180 minutes per session with the average being 60 minutes. The length of the interventions, excluding follow-up, ranged from 4–24 weeks; the median was 12 weeks.

Attrition in FM subjects ranged from 0–67% (median 20%, mean 21%) while controls ranged from 0–48% (median 8%, mean 14%). Compliance was not calculable in the majority of studies. Some studies analyzed data on intent to treat basis and did not report number of sessions subjects attended. Others stated that "the majority" of subjects completed a certain number of classes or that there was a natural break in the data at a certain number of classes. This is problematic in that the "dose" of the intervention was not generally attainable.

The outcome measures in most studies were one or more FM symptoms, measured either on a visual analogue scale, the Fibromyalgia Impact Questionnaire 14 or a health status measure. Fewer also measured fitness markers (strength, flexibility, aerobic capacity). The timing of the measures were pre- vs. post- as compared to multiple time points during the intervention Most failed to explicitly state which outcome was their a priori primary dependent variable. None used real-time symptom monitoring with electronic diaries.

Thirty-nine of the studies were randomized, controlled trials with examiners blinded to treatment allocation. The remaining seven were single group (6 studies), or non-randomly assigned multi-group interventions (1 study). Statistical analyses ranged from questionable paired t-tests, uncorrected for multiple comparisons with no stated a priori hypothesis and within group changes (paired t-tests, change scores and effect sizes) to appropriate statistical methods including independent group t-test, ANOVA and ANCOVA. Abstracts as opposed to full text articles often had inadequate descriptions of methods and analyses making it difficult to confirm the validity of their stated conclusions.

• Most fitness measures improved in people who could tolerate the intervention (e.g., 1 – RM or isokinetic dynamometry strength, time on treadmill, V0₂ max or peak, 6 – minute walk, flexibility).

• The exercise interventions in most studies did not meet the current exercise recommendation for health as developed by the Centers for Disease Control and Prevention and the American College of Sports Medicine 1516 (30 minutes of moderate intensity exercise on most days of the week for health related benefits).

• Those studies that used a higher heart rate or RPE, higher impact movements (e.g., running, jumping) or those where subjects could not self-adjust exercise intensity (e.g. during a flare) suffered the highest attrition rates.

• Subjects attained symptom relief, particularly decreased pain and fatigue as well as improved sleep and mood, with low to moderate intensity exercise of any type. Even very low movement therapies such as QiGong had significant effect sizes for symptom improvement.

• Those studies with 50% maximum heart rate had lower attrition and better symptom improvement than those with the higher intensity.

• Higher intensity studies resulted in greater fitness gains compared to lower intensity in subjects who could complete the intervention.

• Subjects attained symptom relief, particularly decreased pain and fatigue as well as improved sleep and mod, with low to moderate intensity exercise of any type. Even very low movement therapies, such as QiGong, had significant symptom improvement.

• Strength and flexibility training are beneficial for symptom control and fitness improvements but there are insufficient data for recommending a uniform, evidence-based prescription for either of these modalities.

• Descriptive data as well as exercise intervention studies in men, minorities, children and older adults with FM are lacking. The fitness gains in older subjects were comparable to gains seen in age matched healthy controls and were significant compared to the subject's own baseline scores.

• No FM intervention to date has included only overweight or obese persons or individualized the intervention to their unique movement needs (e.g., lower extremity joint protection during weight bearing, awareness of comorbidities such as plantar fasciitis, ankle tendonitis, knee osteoarthritis and a myriad of psychological stigma regarding appearance).

• There is a lack of couples or family based exercise studies in FM, though these are common in healthy elderly, heart disease and other chronic illnesses 171819.

• Determine optimal dosing of exercise so that an evidence based exercise prescription that includes mode, intensity, duration and frequency can be recommended.

• Determine the dose of exercise that effectively manages symptoms versus the dose that produces a symptom flare. This flare is more pronounced that the well documented delayed onset muscle soreness experienced by health deconditioned persons without FM who engage in unfamiliar muscle activity 2021.

• Systematically tract the actual amount of exercise performed compared to the prescribed amount of exercise based on study protocol. Summarize and report these deviations in publication to help identify subgroups of FM patients that are unable to achieve a given level of activity.

• Select uniform symptom and outcome measures for FM exercise trials. Ideally symptoms could be monitored in "real time" rather than retrospectively. This approach would minimize recall bias and allow tracking of symptom trajectory over time. Calling subjects on some type of routine basis or having subjects carry a preprogrammed electronic device that alarms at set intervals requesting real time symptom data would be two ways to do this 2223. Outcome measures should include a patient graded global improvement score as is common in FM medication trials and recommended by the OMERACT 7 workshop 24.

• Examine the combined role of medications and exercise. Many FM subjects take medications and are told to exercise, yet only three studies thus far compare the combined and separate roles of exercise plus specific medications in FM 252627, although many more acute dosing/cross sectional trials of drug and exercise in FM have been reported. At minimum, medications should be monitored and their use considered in statistical analyses.

• Include cost-utility analysis of exercise as a treatment for FM in future trials.

• Integrate families or other support systems into lifestyle interventions such as exercise as a way of improving long-term compliance.

• Test exercise modalities and movement therapies for a broader array of physical and mental health outcomes, beyond symptoms and physical fitness. For example, descriptive studies have found deficits in balance and increased falls in FM patients 282930, yet only one intervention study measured balance as an outcome 54.

• Maximize methodological rigor. Randomization should be applied whenever possible to equally distribute variance throughout the groups. Hypotheses should be stated a priori and tested with appropriate correction for multiple comparisons and covaried for baseline differences between groups. CONSORT guidelines for reporting findings should be followed 31.

• Report compliance by calculating the number of classes or minutes attended divided by the number offered. Reporting compliance is critical as it allows reviewers to calculate the "dose" of the intervention that the subject actually received, similar to a pill count in a medication study.

• Conduct larger, longer-lasting RCTs that follow the individual from low impact exercise (e.g. pool settings to group based, land laboratory settings to home exercise with weekly booster sessions in community-based venues. This approach would better simulate a real-world application of exercise.

• Evaluate methods to increase compliance in longer trials to test techniques such as motivational interviewing.