Treatment of fibromyalgia with cyclobenzaprine: A meta-analysis


  • Jeanne K. Tofferi,

    1. Walter Reed Army Medical Center, Washington, DC and Uniformed Services University of the Health Sciences, Bethesda, MD
    Search for more papers by this author
    • Presented by Dr. Tofferi in partial fulfillment of the requirements for a Masters of Public Health degree, Uniformed Services University of the Health Sciences, Bethesda, MD.

    • The opinions expressed in this article are those of the authors and should not be construed in any way to reflect those of the U.S. Army or the Department of Defense.

  • Jeffrey L. Jackson,

    Corresponding author
    1. Uniformed Services University of the Health Sciences, Bethesda, MD
    • Uniformed Services University of the Health Sciences, Department of Medicine (EDP), 4301 Jones Bridge Road, Bethesda, MD 20814–4799===

    Search for more papers by this author
    • The opinions expressed in this article are those of the authors and should not be construed in any way to reflect those of the U.S. Army or the Department of Defense.

  • Patrick G. O'Malley

    1. Walter Reed Army Medical Center, Washington, DC and Uniformed Services University of the Health Sciences, Bethesda, MD
    Search for more papers by this author
    • The opinions expressed in this article are those of the authors and should not be construed in any way to reflect those of the U.S. Army or the Department of Defense.



To systematically review the effectiveness of cyclobenzaprine in the treatment of fibromyalgia.


Articles describing randomized, placebo-controlled trials of cyclobenzaprine in people with fibromyalgia were obtained from Medline, EMBase, Psyclit, the Cochrane Library, and Federal Research in Progress Database. Unpublished literature and bibliographies were also reviewed. Outcomes, including global improvement, treatment effects on pain, fatigue, sleep, and tender points over time, were abstracted.


Five randomized, placebo-controlled trials were identified. The odds ratio for global improvement with therapy was 3.0 (95% confidence interval [95% CI] 1.6–5.6) with a pooled risk difference of 0.21 (95% CI 0.09–0.34), which calculates to 4.8 (95% CI 3.0–11) individuals needing treatment for 1 patient to experience symptom improvement. Pain improved early on, but there was no improvement in fatigue or tender points at any time.


Cyclobenzaprine-treated patients were 3 times as likely to report overall improvement and to report moderate reductions in individual symptoms, particularly sleep.


Fibromyalgia is a chronic musculoskeletal pain disorder affecting up to 2% of the United States population (3.4% women and 0.5% men). The prevalence increases with age, present in up to 7% of women in their seventh and eighth decades (1). Fibromyalgia is common in primary care clinics and is second only to rheumatoid arthritis as a cause for visits to rheumatologists (2). Unfortunately, treatment options for fibromyalgia patients remain limited both in number and effectiveness.

Fibromyalgia is typified by diffuse muscle pain and aches, muscle stiffness, sleep disturbance, fatigue, and tenderness over discrete tender points. The cause and pathophysiology of this syndrome remain unclear. Treatment trials have included both pharmacologic and nonpharmacologic therapies with mixed results. A few have demonstrated some improvements in fibromyalgia symptoms over the short term. This improvement is modest however, affecting only a minority of study patients.

Drug treatment options for fibromyalgia include nonsteroidal antiinflammatory drugs (NSAIDs) (3, 4), tricyclic antidepressants (3, 5–8), other antidepressants including selective serotonin reuptake inhibitors (9–12), cyclobenzaprine (6, 13–17), S-adenosylmethionine (SAM-e) (18), and 5-hydroxytryptophan (19). It appears that the most effective agents affect neurotransmitters at the receptor site (20). Antidepressants have been shown to be have some effect, whereas NSAIDs (3, 4), steroids (21), and calcitonin (22) have not. Cyclobenzaprine (6, 13–17), alprazolam (23), tramadol (24), and SAM-e (18) have shown some benefit to patients in small studies.

Nonpharmacologic treatment modalities, including exercise, physical therapy, acupuncture, and alternative medicines, have shown some benefit in selected patients, but their roles remain unclear and are less frequently offered (20). Despite the various treatment modalities available, the vast majority of fibromyalgia patients remain in significant pain years after their original diagnosis (25), with the majority (75%) requiring medication to control symptoms (26).

Cyclobenzaprine, a muscle relaxant that modulates muscle tension at a supraspinal level (27), is used in up to 12% of patients diagnosed with fibromyalgia (28). Our purpose was to conduct a systematic review to assess the magnitude and duration of effect of cyclobenzaprine in the treatment of fibromyalgia.


We searched Medline (1966 to November 2000), Psyclit (1974 to November 2000), and EMBase (1974 to November 2000) using the following key words (all languages): fibrositis, fibromyalgia, cyclobenzaprine, flexeril, and chronic fatigue disorder. We searched the Cochrane Database of Systematic Reviews, the Database of Abstracts of Reviews of Effects, and the Cochrane Central Register of Controlled Trials. We also searched the Federal Research in Progress Database to identify unpublished literature. We contacted Merck, the manufacturer of Flexeril, for any unpublished data. We researched references of reviewed articles for additional articles missed by the computerized search. All primary articles were reviewed independently in duplicate.

Studies were screened for inclusion through review of the published article, based on the following criteria: randomization, placebo control, 1 group treated with cyclobenzaprine, and measurable outcomes. Each article was reviewed for inclusion by 2 investigators.

Study quality was assessed using the 6-item instrument developed by Jadad et al (29). Study quality was assessed independently, with substantial interrater agreement (intraclass correlation coefficient = 0.97). Disagreements were arbitrated by consensus.

Abstracted data included treatment characteristics (dose, duration of therapy), country of origin, number and sex of participants, adverse events, losses, and outcomes. Outcomes were recorded as either dichotomous or continuous variables based on how they were reported in the original study.

Analyses were done using Stata software (30). Assessment for publication bias was done using the methods of Egger et al (31) and Begg and Mazumdat (32). Heterogeneity was assessed with the chi-squared statistics using the methods of Ysuf et al (33). Although pooled results were homogenous, a random effects model using the method of DerSimonian and Laird (34) was used to calculate the summary odds ratio, risk difference, and standardized mean difference (SMD) due to concern for inherent heterogeneity among drug doses, schedules, and study designs. Analysis of continuous outcomes involved comparing standardized differences in means between the control and treatment groups. Means for the 4 continuous outcomes (pain severity, fatigue, number of tender points, and sleep quality) for both placebo and treatment groups were standardized by dividing by their standard deviations. The difference between these standardized means were calculated for each study for which it was available and analyzed. By standardization, the study results were transformed to a common scale (standard deviation units) that facilitated pooling. By convention, an effect size of 0.2 is considered small, 0.5 is moderate, and >0.8 is considered large (35). The beneficial effects of placebo were also assessed against baseline data to determine the significance of this effect.

For every significant test result, we tested the sensitivity of our results to potential unpublished studies using a file drawer test (36). This test determines how many negative studies (odds ratio = 1.0, standardized mean difference = 0), with variance and size equal to the average seen among the included studies, would be needed to negate our findings. We also tested for the relative influence of the year of publication, the country of origin, and study quality scores on the results.


Our literature search identified 27 citations involving fibromyalgia or fibrositis and flexeril or cyclobenzaprine. Six articles met initial inclusion criteria (6, 13–17). Of the 6 randomized trials, 1 was excluded because it had no true placebo arm (17).

Qualitative synthesis.

Details of individual studies are presented in Table 1. Although most were conducted prior to the publication of the American College of Rheumatology (ACR) criteria for fibromyalgia in 1990 (37), inclusion criteria were remarkably consistent. All studies included the number of tender points and widespread pain, consistent with the current ACR criteria. Other criteria included nonrestorative sleep, widespread pain in 3 or more body areas lasting a minimum of 3 months, absence of secondary causes, and exacerbation of symptoms by physical or emotional stress.

Table 1. Characteristics of included studies*
Author, country, yearInclusion criteriaDrug, dosage, sample sizeLength of trial, weeksQuality score, 0–8Quality problems
  • *

    ACR = American College of Rheumatology; VAS = visual analog scale.

  • Minor criteria: modulation of symptoms by physical activity, weather, anxiety, stress, poor sleep, fatigue/tiredness, chronic headache, irritable bowel syndrome, subjective swelling, or numbness.

Carrette, Canada, 1994ACR criteria, >4 of 10 on a pain VAS, and fibromyalgia symptomsCyclobenzaprine, 10 mg every evening titrated to 10 mg every morning and 20 mg every evening based on patient tolerance, n = 82; versus amitriptyline, n = 84; versus placebo, n = 4224 (parallel design)8Inadequate discussion of adverse events, placebo not described as identical, withdrawals >10%
Reynolds, Canada, 1991Seven tender points, nonrestorative sleep, fatigue, and absence of biochemical or serologic evidence of another diseaseCyclobenzaprine, 10 mg 3 times daily, n = 12; versus placebo, n = 124 (crossover design)3Inadequate description of randomization, placebo not described as identical, inadequate description of selection and exclusion criteria, withdrawals >10%, no intention-to-treat analysis
Hamaty, US, 1989Generalized aching pain in ≥3 body areas for ≥3 months, 8 tender points, insomnia or nonrestorative sleep, and normal lab resultsCyclobenzaprine, 10–40 mg patient-driven dosing, n = 11; versus placebo, n = 118 (crossover design)1No intention-to-treat analysis, no analysis of losses, inadequate description of randomization, placebo not described, no clear description of inclusion and exclusion criteria, withdrawals >10%, no sample size calculations
Quimby, US, 1989Aches, pain, and stiffness in ≥3 sites for ≥3 months; normal lab and radiograph results; ≥5 tender points; and 3 minor criteria†Cyclobenzaprine, 10 mg every evening titrated up to 10 mg every morning and 30 mg every evening, n = 61; versus placebo, n = 206 (parallel design)6Inadequate description of randomization, withdrawals >10%, no intention-to-treat analysis, no sample size calculations
Bennett, US, 1988Widespread musculoskeletal pain for ≥3 months, 7 tender points, morning stiffness, sleep disturbance, and 2 minor criteriaCyclobenzaprine, 10 mg every day titrated up to 40 mg as needed, n = 61; versus placebo, n = 5812 (parallel design)6Inadequate description of randomization, withdrawals >10%, no sample size calculations

The 5 randomized, controlled trials included a total of 312 patients. The median number of patients in each study was 62 (range 56–69); the majority were women (95%), averaging 46 years in age. The median study duration was 6 weeks (range 2–24 weeks) and the average symptom duration among participants was 39 weeks (range 4.5–132 weeks). Studies generally included a once-daily dose of cyclobenzaprine or placebo. Studies that used cyclobenzaprine dosages of 30 mg/day or greater split the doses, with a smaller (10 mg) morning dose and a larger bedtime dose (20 mg or greater). Dosages were generally started at 10 mg at bedtime and titrated as tolerated. If participants found the medication to be too sedating, the dosage (placebo or cyclobenzaprine) was changed to the previous dosage, with a minimum of 10 mg cyclobenzaprine or “equivalent” placebo.

Abstracted outcomes included global improvement (abstracted as a dichotomous variable) and treatment effects on pain, fatigue, sleep, and tender points (number and/or severity). Treatment effects were extracted as continuous variables at 3 time points (4, 8, and 12 weeks). Although all studies included measures of sleep and pain, not all the other outcomes were extractable from all studies.

Overall, the quality of the studies was fair, with an average quality score of 4.4 (range 0–8), similar to the quality of other studies in fibromyalgia (8). Problems with the studies included no statement of the method of randomization, failure to identify blinding (6, 13–16), lack of a clear description of exclusion criteria (13, 15), no method to assess adverse effects (13, 15), inadequate description of losses (13), no intention-to-treat analysis (14–16), and large losses to followup (6, 13, 15, 16). On average, only 71% of cyclobenzaprine patients and 57% of placebo patients completed the studies.

Quantitative synthesis.

Global improvement was extractable from 3 studies as dichotomous data. Overall, patients treated with cyclobenzaprine were more likely to report themselves to be “improved” (odds ratio [OR] 3.0, 95% confidence interval [95% CI] 1.6–5.6; Figure 1) than those treated with placebo. These results were homogenous (P = 0.53; but with evidence of publication bias, P = 0.05). The pooled risk difference was 0.21 (95% CI 0.09–0.34), which is equivalent to 4.8 patients (95% CI 3.0–11) needing treatment for 1 patient to experience symptom improvement.

Figure 1.

Individual study and summary effect size on the dichotomous outcomes of improvement. 95% CI = 95% confidence interval.

Continuous outcome data (sleep, pain, fatigue, tender points) were extracted at 4, 8, and 12 weeks. In each of these time periods, 2–3 studies reported data on each variable. A statistically significant improvement in sleep was noted for all time points with SMDs ranging from 0.34 to 0.50 (Figure 2). Pain was improved in the cyclobenzaprine group in week 4 only (SMD = 0.35). No improvement in fatigue or tender points was noted at any time point. These analyses were homogenous for fatigue, sleep, pain, and tender points at all time points. Most continuous outcomes had no evidence of publication bias, with the exception of pain at 4 weeks (P = 0.02).

Figure 2.

Summary standardized mean differences (SMD) of sleep, pain, fatigue, and tender points at 4, 8, and 12 weeks. Tender pt = tender points.

To assess the possibility of a placebo effect, the SMDs for the placebo arms were calculated for each of the continuous variables for the 3 time points and compared with the patients' baseline values. Statistically significant improvement in sleep was noted at all 3 time points with SMDs ranging from 0.42 to 0.52 (Figure 3). A statistically significant improvement in pain was noted at 12 weeks only (SMD = 0.39) in the placebo group. No improvement in fatigue or tender points in the placebo arm was noted at any time point over baseline.

Figure 3.

Summary measure of improvement over baseline for sleep and pain in patients treated with placebo at 4, 8, and 12 weeks. SMD = standardized mean difference.

Sensitivity analysis.

Our results were not overly influenced by any single study. There was no evidence of a differential effect, using metaregression, when controlling for the quality score, year of publication, percentage of women patients, age of patients, and symptom duration.

File drawer tests were conducted for each significant result. For overall global improvement, the number of unpublished studies needed to negate the positive effect of cyclobenzaprine was 38. To negate the improvement seen in individual symptoms, the number of unpublished negative studies needed ranged from 14 to 45.


Metaanalysis of these 5 published, fair-quality studies suggests that cyclobenzaprine offers some benefit to patients with fibromyalgia. Patients are about 3 times as likely to report improvement in their symptoms, and 5 patients would need to be treated for 1 to improve. The average treated patient would experience a moderate amount of improvement in sleep at all time points, pain would be modestly improved, but only early on, and patients would experience no improvement in fatigue or tender points. Comparing our results with that of antidepressants in another metaanalysis, antidepressants appear to provide somewhat more benefit with an overall OR of 4.2 (95% CI 2.6–6.8). Methodologic differences in these 2 studies prevent further comparison (8).

It is difficult to say whether these beneficial effects would wane over time. Although Figure 2 would suggest the benefit from treatment diminishes from 4 to 12 weeks, there is not enough evidence to report that the benefit of cyclobenzaprine in fibromyalgia decreases with time.

For the placebo arms, there is a modest improvement in sleep at all time points measured, which is fairly consistent throughout the time points and does not appear to diminish. Clearly, studies of interventions with fibromyalgia need to include a placebo arm to ensure delineation of treatment from placebo effects.

There are several weaknesses of this meta-analysis. All studies were marred by a high dropout rate and only 2 of the studies included an intention-to-treat analysis. Another problem in this group of studies was difficulty blinding patients to treatment. Eighty-five percent of patients taking cyclobenzaprine experienced side effects. One study asked patients to guess whether they were on active drug or placebo. Patients receiving cyclobenzaprine were right 65% of the time (14).

The short duration of each of these studies is also problematic. Although they all demonstrated a benefit over the course of their study period, it is difficult to extrapolate the results of short studies to a chronic condition. Finally, a crossover design was used for 2 of the 5 included studies. Although cyclobenzaprine is not known for its long-lasting effects, a carryover effect may have persisted after the washout period, thereby masking a larger treatment effect. The crossover design is particularly problematic when a large percentage of patients are able to guess whether they are receiving active treatment.

This metaanalysis of 5 published, randomized controlled trials suggests that cyclobenzaprine improves global functioning of patients with fibromyalgia with a modest improvement in sleep quality. There may also be a small improvement in patient reports of pain, but there appears to be no effect on fatigue or tender points. To clarify the true benefit of cyclobenzaprine in fibromyalgia, future studies would be of longer duration (months to years), include larger numbers of patients, with more complete followup and more rigorous attempts at blinding of patients and investigators.