PubMed Health. A service of the National Library of Medicine, National Institutes of Health.

Smith B, Peterson K, Fu R, et al. Drug Class Review: Drugs for Fibromyalgia: Final Original Report [Internet]. Portland (OR): Oregon Health & Science University; 2011 Apr.

Results

Overview

Literature searches identified 1148 citations. We received dossiers from 2 pharmaceutical manufacturers: Eli Lilly and Company and Forest Laboratories Inc. By applying the eligibility and exclusion criteria to titles and abstracts of all identified citations, we obtained full-text copies of 119 citations. After re-applying the criteria for inclusion, we ultimately included 51 publications. Five included studies were identified after expanding the population inclusion criteria to include a broadened definition of fibromyalgia or fibrositis.38–42 See Appendix E for a list of excluded studies and reasons for exclusion at this stage. Figure 1 shows the flow of study selection.

Figure 1. Results of literature search.

Figure 1

Results of literature search.

Key Question 1. For adults with fibromyalgia, what is the comparative effectiveness/efficacy of included interventions?
1a. When used as monotherapy?
1b. When used as adjunctive therapy?

Summary of Findings

Direct evidence

  • There was low-strength evidence that immediate-release paroxetine is superior to amitriptyline in reducing pain (−28% compared with − 1%; z= − 5.64; P<0.001) and sleep disturbance (− 39% compared with − 13%; z= − 4.62; P<0.001) based on an 100-mm visual analog scale, but there were no differences in fatigue, tender points, depression, or in the numbers of patients with moderate or marked improvement based on clinical global assessment
  • There was low-strength evidence of no significant differences between amitriptyline and cyclobenzaprine or nortriptyline in any efficacy outcomes.

Indirect evidence

Pooled analysis
Indirect meta-analysis
Comparisons to placebo
  • Gabapentin significantly improved pain severity and response, overall impact of fibromyalgia, global status, and sleep, but not tender point pain threshold, depression or overall quality of life
  • Compared with placebo, a significant reduction in pain severity was only found with cyclobenzaprine in 1 of 3 trials
  • Among selective serotonin reuptake inhibitors, only fluoxetine, at a higher dose (45 mg), resulted in significantly greater improvements than placebo in pain, fatigue, and Fibromyalgia Impact Questionnaire Total Score
  • Controlled-release paroxetine did not significantly decrease pain, disability, or depressiveness or increase the number of patients with a 50% or greater response, but did significantly decrease the Fibromyalgia Impact Questionnaire Total Score, fatigue, and improved global status
  • Citalopram did not significantly improve pain or fatigue and only reduced depression and improved sleep in 1 of 2 trials.

Detailed Assessment

Direct evidence

Direct evidence regarding the comparative effectiveness among included interventions was limited and only available from 4 small randomized controlled trials that compared amitriptyline to cyclobenzaprine (N=208),43 fluoxetine (N=31),44 nortriptyline (N=118),45 and immediate-release paroxetine (N=68).46 All patients met the American College of Rheumatology 1990 criteria for classification of fibromyalgia. Three trials reported duration of fibromyalgia, which ranged from 36 months46 to 101 months.43 Participants were 95% female with mean ages ranging from 36 years46 to 53.4 years.45 Race was 100% Caucasian in the fluoxetine trial conducted in Massachusetts,44 62% Caucasian and 38% non -Caucasian in the Brazilian trial of nortriptyline,45 and was not reported in the Canadian study of cyclobenzaprine43 or the Turkish study of immediate-release paroxetine.46 Trial settings included outpatient rheumatology clinics43, 45 and a tertiary referral center,44 but was not well described in the Turkish study.46 Trial durations ranged from 6 weeks44, 46 to 6 months.43 Amitriptyline dosage ranged from 25 mg 44, 45 to 100 mg.46 Mean dosages for the comparator drugs were 20 mg for cyclobenzaprine, 20 mg for fluoxetine, 20 mg for immediate-release paroxetine, and 25 mg for nortriptyline.

Three trials were rated fair quality 43, 45, 46 and the third was rated poor quality.44 The main limitation of the poor-quality trial was that its analyses excluded a large proportion of the data – over one-third; consequently, its results will not be discussed here, but can be found in Evidence Tables 1 and 2.44

Only 146 of 3 trials43, 45, 46 found any significant differences between drugs in efficacy, in which there was low-strength evidence that immediate-release paroxetine 20 mg was superior to amitriptyline 100 mg over 6 weeks in reducing pain (− 28% compared with − 1%; z= −5.64; P<0.001) and sleep problems (− 39% compared with − 13%; z= − 4.62; P<0.001), as measured based on a 100-mm visual analog scale.46 However, differences between immediate-release paroxetine and amitriptyline were not significant for change in fatigue (− 9% compared with −5%; z=0.74), tender points, depression, or in the numbers of patients with moderate or marked improvement based on clinical global assessment (22% compared with 17%; relative risk, 1.27; 95% CI, 0.47 to 3.47).

Otherwise, the remaining 2 trials provided low-strength evidence of no significant differences between amitriptyline and cyclobenzaprine43 or nortriptyline 45 in any efficacy outcomes. The 2 head-to-head trials were inconsistent in their methods for assessing all efficacy outcomes. Compared with amitriptyline, similar reductions were found for cyclobenzaprine in visual analog scores for pain (− 33% compared with − 28%) and fatigue (− 33% compared with − 32%), McGill Pain Questionnaire Pain Rating Index scores (− 32% compared with − 31%), Depression Scale scores from the Arthritis Impact Measurement Scale (− 25% compared with − 20%), and in the Health Assessment Questionnaire Disability Index (− 15% for both drugs). Similar proportions of patients were classified as responders based on meeting at least 4 of the following 6 criteria: 50% improvement in pain, sleep, fatigue, patient global assessment, or physician global assessment, and increase of 1 kg in mean total myalgic score (33% compared with 36%).43 Also compared with amitriptyline, similar reductions were found for nortriptyline in mean number of tender points (− 3 compared with − 2.2) and in the Fibromyalgia Impact Questionnaire total score (27% compared with 37%), and similar proportions of patients met criteria for slight to great improvement based on the Verbal Evaluation Scale for Global Improvement (72% compared with 86%).45

We found 5 systematic reviews that assessed multiple drugs for the treatment of fibromyalgia.47–51 Four reviews were graded as good quality47–50 and 1 was graded as fair quality.47 Only 1 of these reviews performed an indirect meta-analysis to compare the effectiveness between duloxetine, milnacipran, and pregabalin in fibromyalgia from placebo-controlled trials of individual drugs.49 Although it was a recent review, we chose to update the indirect meta-analysis to include evidence from 2 new trials for milnacipran that are now available.52, 53 We have reported results from the Hauser, et al. analysis where there was no new data or the new data did not change the results.

The remaining reviews had limited usefulness for this report as none provided any additional direct or indirect comparative data. One systematic review reported evidence on antidepressant drugs used for the treatment of fibromyalgia.51 No meta -analysis was performed and the only individual drug data reported was the means across studies for amitriptyline. We conducted a meta-analysis comparing amitriptyline to other drugs to treat fibromyalgia so did not use the data reported in this review.51 One systematic review reported a calculated effect size for amitriptyline and tramadol but only a class effect for antidepressants.47 Given the new criteria for diagnosis of fibromyalgia released in 2010, we extended our inclusion criteria and included earlier studies of fibromyalgia not included in their review so did not report on their outcomes. One systematic review of amitriptyline reported on 10 randomized controlled trials but did not perform a meta-analysis due to large clinical variability and statistical heterogeneity.50 The results are considered as appropriate below. One systematic review pooled data of pregabalin and gabapentin and reported on a class effect size rather than comparative data which limited its usefulness for this report.48

Indirect evidence

Indirect meta-analysis

Of the 8 drugs included in this review, we found that only 4 drugs had multiple trials with sample sizes adequate to perform a comparative analysis. All trials used the drugs as monotherapy and no trial evaluated the effectiveness of the drugs as adjunctive therapy. We performed a meta-analysis of 6 placebo -controlled trials of amitriptyline,39, 43, 45, 54–564 placebo -controlled trials of pregabalin,57–60 5 placebo -controlled trials of milnacipran,52, 53, 61–64 and 4 placebo-controlled trials of duloxetine.65–69 Two trials of amitriptyline were identified by expanding our definition of fibromyalgia to include criteria for fibrositis that would fall under the umbrella of the updated definition of fibromyalia in 2010.39, 70 Data from Scudds, et al. was excluded from the 8–15 week meta-analysis due to its short duration of 4 weeks. One additional trial of pregabalin was included but did not contribute data to our analysis due to significant methodological variance from the other trials and given that its outcome was loss of effectiveness in responders. The trial results are reported below. 71 No trials were of long duration (8 to 28 weeks). Sample size was small for all of the amitriptyline trials (N=22 to 126) but generally moderate to large for the rest (N=125 to 1025). All of the trials were rated as fair quality. The trials of duloxetine, milnacipran, and pregabalin were industry sponsored whereas only 1 of the amitriptyline trials were industry sponsored.56 The baseline demographics for all of the trials were similar, with the majority of patients being middle aged (mean age range 40–53 years), white (65% to 93%), and female (82% to 100%). For pregabalin, we excluded data on the lowest dose of 150 mg once daily as only 1 trial included this dose, N=131, and a prior meta-analysis found this dose to be no different than placebo.58, 72 Otherwise we combined data on different daily dosage groups including amitriptyline 25–50 mg, pregabalin 300–600 mg, milnacipran 100–200 mg, and duloxetine 40–120 mg. We reported on outcomes of clinical importance and with adequate data to perform the analysis. We performed our analysis using short-term data (8–15 weeks) as this was consistent across most studies. We performed a sensitivity analysis to determine if differences in outcome were seen using long-term trials. We excluded trials shorter than 8 weeks from our analysis.42, 44, 73 The results and quality of these trials are summarized in Evidence Tables 1 and 2.

Pain

Multiple measures of pain were used across studies and at variable intervals. We performed our analysis based on a 10-point derivative scale as this was most consistently reported across trials and is considered a valid method of pain assessment. All drugs were effective in improving average pain or 24 hour recall of pain compared with placebo, with amitriptyline showing the most effectiveness (− 1.68;95% CI, − 3.3 to − 0.05), followed by duloxetine (− 0.94; 95% CI, − 1.3 to − 0.61), pregabalin (− 0.67; 95% CI, − 0.94 to − 0.40), and milnacipran (− 0.49; 95%CI, − 0.65 to − 0.33) (Table 3). Significant heterogeneity was seen with the amitriptyline (I2=72.7%) and pregabalin trials (I2=62.6%). Given that the Ginsberg 1996 trial used a sustained-release formulation of amitriptyline, we repeated the analysis excluding the data from this trial and found that although the result was in the direction of improvement, the significance for amitriptyline was lost with a pooled mean difference of − 0.94 (95%CI, − 2.18 to 0.31), I2=58.3%. A similar effect was seen when we repeated the analysis excluding the data from Carette 1986 which did not use the American College of Rheumatology criteria for population inclusion, with a pooled mean difference of − 1.82 (95% CI, − 3.99 to 0.35), I2=81%. Nishishinya, et al. performed a systematic review of amitriptyline and identified 10 placebo-controlled trials studying the effectiveness of amitriptyline in treating fibromyalgia.50 They did not perform a pooled analysis due to clinical and statistical heterogeneity but did find significant improvement in pain for the trials of amitriptyline 25 mg compared with placebo, consistent with our analysis.50

Table 3. Pooled effectiveness of amitriptyline, pregabalin, milnacipran, and duloxetine compared with placebo (8–15 weeks).

Table 3

Pooled effectiveness of amitriptyline, pregabalin, milnacipran, and duloxetine compared with placebo (8–15 weeks).

Indirect meta-analysis of all placebo-controlled trials found that there was no difference between the drugs except that duloxetine was superior to milnacipran (difference in mean difference, − 0.45;95% CI, − 0.80 to − 0.08). This finding held true when the analysis was repeated, excluding the Ginsberg 1996 and the Carette 1986 trial data for amitriptyline. The recent meta-analysis performed by Hauser49 was consistent with the finding that duloxetine was superior to milnacipran but unlike our analysis, they found that duloxetine and milnacipran were also superior to pregabalin (standardized mean difference, 1.22; 95%CI, 1.10 to 1.80 and standardized mean difference, 0.70; 95% CI, 0.58 to 0.82, respectively). One of the limitations of the trials was that there were multiple ways in which pain was reported in each trial, including daily and weekly symptoms as well as multiple different pain scales between trials. We analyzed average or 24-hour daily pain score, converting it to a 0–10 scale, and restricted our analysis to data 8–15 weeks in duration. It is unclear how Hauser combined the multiple reports of pain. Additionally, the median duration of the randomized phase of the trials from the Hauser meta-analysis was 24 (range 6–28) weeks.49 When we repeated our analysis including all trial data (duration range 6–28 weeks), the superiority of duloxetine over milnacipran was lost (−0.32; 95% CI, − 0.68 to 0.04). In summary, for short-term use, there was low evidence that all drugs are superior to placebo in pain response, with no difference between the drugs except that duloxetine was superior to milnacipran. Head-to-head trials are needed to confirm these findings.

Response

Response was defined differently in all of the trials, with many of the trials having a composite response that included multiple outcome measures such as 30% or 50% reduction in pain and improvement on Patient Global Impression of Improvement or Change. Many of the trials also reported this data separately, allowing us to perform a pooled analysis and indirect meta-analysis on pain response rate. Pooled analysis of placebo-controlled trials found that all drugs were effective in achieving a 50% improvement in pain (Table 4, Figure 2). The data for amitriptyline was insufficient to determine its validity given the small number of patients from 2 trials, N=68, one of which used a sustained release formulation of amitriptyline.54, 55 Two systematic reviews of pregabalin compared with placebo in fibromyalgia used the same 4 trials, pooled data based on dose, and found similar results on 30% or 50% pain response and for “much or very much improved” on Patient Global Impression of Improvement or Change.72, 74 Indirect meta -analysis of the placebo-controlled trials of duloxetine, milnacipran, and pregabalin found that there was no significant difference between the drugs on ability to achieve a 50% reduction in pain(T able 4). The data for amitriptyline was too sparse for indirect comparison. We had no data on amitriptyline for the outcome of 30% improvement in pain but all of the other drugs had a small but significant improvement in 30% pain response compared with placebo and no significant difference between the drugs was found (Tables 3 and 4). This finding was consistent with the Hauser, et al. analysis.49

Table 4. Indirect analysis of placebo-controlled trials in fibromyalgia.

Table 4

Indirect analysis of placebo-controlled trials in fibromyalgia.

Figure 2. Response rate 50% improvement in pain.

Figure 2

Response rate 50% improvement in pain.

For the outcome of any improvement on the Patient Global Impression of Improvement or Change score, a significant benefit over placebo was found for milnacipran, pregabalin, and amitriptyline. However, I2 statistics revealed substantial heterogeneity within the milnacipran (74.4%) and pregabalin (73.6%) trials. When we used meta-regression to explore reasons for the heterogeneity, we found a significant association for placebo group response rate (P=0.008), but not for definition of improvement (any compared with much or very much). Findings from the meta-regression indicated that trials with higher rates of improvement in the placebo group had smaller benefits with milnacipran or pregabalin. In our indirect meta-analysis, no significant differences were found between milnacipran, pregabalin, and amitriptyline. However, the data on amitriptyline was insufficient to make any conclusions given that there was only 1 small trial that reported on this outcome (N=80). When we repeated our analysis considering only those who reported much or very much improvement on the Patient Global Impression of Improvement or Change score, no difference was found between milnacipran and pregabalin (ratio of relative risk, 0.85; 95% CI, 0.59 to 1.24)(T able 4). Trials of duloxetine that reported on this outcome provided only the mean change rather than the absolute number of patients reporting improvement and thus could not be included in our analysis. Arnold, et al. performed a pooled analysis of the same 4 duloxetine trials and reported a significant difference between duloxetine and placebo with 38.4% of duloxetine-treated patients reporting at least much improved compared with 21.7% of placebo patients, P<0.001.75

One additional trial of pregabalin was reviewed but not included in our analysis due to significant methodological variance from the other trials.71 In a 26-week placebo-controlled randomized discontinuation trial of patients with fibromyalgia who had achieved at least 50% reduction on the visual analogue scale and much or very much improvement on the Patient Global Impression of Improvement or Change score after a 6-week open-label treatment period (n=566), the time to loss of response (<30% reduction in pain) was longer for pregabalin than for placebo (34 days compared with 7 days; P<0.0001).71

In summary, all drugs were superior to placebo in 50% response rate and Patient Global Impression of Improvement or Change score. There was low evidence that no differences exist between pregabalin, duloxetine, or milnacipran on pain response rate with insufficient evidence to report on this outcome for amitriptyline. There was insufficient evidence to draw conclusions of the comparative effectiveness on the Patient Global Impression of Improvement or Change score.

Fatigue

Fatigue was defined differently between the drugs. We considered the general fatigue score of the multidimensional fatigue inventory, the fatigue score of the Fibromyalgia Impact Questionnaire, the Visual Analogue Scale, and the global fatigue index of the Multidimensional Assessment of Fatigue score, and determined a standardized mean difference based on available short-term data (8–15 weeks). Milnacipran, pregabalin, and amitriptyline were superior to placebo in short-term trials of 8–15 weeks, but not in longer-term trials of 24–28 weeks. There was high heterogeneity noted in the pregabalin trials (I2=62.8%). In their meta-analysis of the same 4 pregabalin trials, Straube, et al. found pregabalin superior to placebo at doses of 300 mg (P<0.01) and 450 mg (P<0.01), but not at 600 mg.74 There was no difference between duloxetine and placebo, with 2 trials reporting on this outcome (N=727). No difference was found between the drugs (Table 3). Our result contradicts the meta-analysis of Hauser, et al. that found milnacipran superior to duloxetine (standardized mean difference, 0.77; 95% CI, 0.67 to 0.87), and pregabalin superior to duloxetine (standardized mean difference, 0.62; 95% CI, 0.52 to 0.72) and milnacipran (standardized mean difference, 0.81; 95% CI, 0.71 to 0.91).49 The difference between our findings and those of Hauser, et al. could be mainly due to the difference in indirect meta-analytic methods. We calculated an absolute difference of standardized mean differences between interventions whereas the Hauser analysis calculated a ratio of standardized mean differences between the drugs, which produces an estimate of the relative effect rather than an absolute difference. The ratio for standardized mean differences is rarely used. When we calculated a ratio of standardized mean differences, we could not replicate the significant value reported by Hauser, et al. We obtained 0.73 (95% CI, 0.13 to 4.19), which indicated that the difference is not significant. Given that the difference is actually small, an analysis based on absolute difference is not significant. In summary, there was low evidence that milnacipran, pregabalin, and amitriptyline are superior to placebo on measures of fatigue, and no differences existed between the drugs.

Function

The Fibromyalgia Impact Questionnaire total score was used to assess overall change in fibromyalgia symptoms and their impact on daily function. It is an instrument designed to reflect the multidimensionality of fibromyalgia by questioning patients about the extent of their symptoms and the effect of these on their activities of daily living.76 On the measure of total score on the Fibromyalgia Impact Questionnaire, mean change from baseline reached the minimally clinically important difference level proposed by Bennett, et al. in all treatment and placebo groups.77 Pooled analysis of short-term (8–15 weeks) placebo-controlled trials of duloxetine, milnacipran, and pregabalin found a significant improvement compared with placebo for all drugs (I2=23.1%) (Table 3). Our results for pregabalin were in agreement with the pooled analysis by Straube, et al. who found pregabalin superior to placebo at all 3 doses (300 mg [P<0.05], 450 mg [P<0.001], and 600 mg [P<0.01]).74 Indirect meta-analysis of placebo-controlled trials of duloxetine, milnacipran, and pregabalin found no significant difference between the drugs (Table 4). The data on amitriptyline was insufficient to make a statement on this outcome as only 1 small trial, N=80, reported data on Fibromyalgia Impact Questionnaire total score, although it did not find a significant difference compared with placebo.45

We also considered the Medical Outcomes Study 36-item Short-Form Health Survey physical and mental component summaries to assess therapeutic response of the included drugs on overall physical and mental function. No differences were found in mean differences between duloxetine, milnacipran or pregabalin on either of these measures based on 6 trials that reported data.52, 53, 57, 61, 65, 68, 69 The pooled analysis of placebo-controlled trials (8–15 weeks) of milnacipran and duloxetine found a small but significant mean difference compared to placebo for the Medical Outcomes Study 36-item Short-Form mental component summary (milnacipran 1.47; 95% CI, 0.74 to 2.19 and duloxetine 0.27; 95% CI, 0.95 to 4.44), whereas only milnacipran found a significant mean difference compared to placebo on the physical component summary (1.08; 95% CI, 0.54 to 1.61)52, 53, 61, 63 When longer-term data (6–28 weeks) were added to the analysis, milnacipran was still found to be significantly superior to placebo on the physical component summary score (standard mean difference, 1.06; 95% CI, 0.52 to 1.60) and no significant benefits were found for duloxetine or pregabalin. In the pooled analysis of pregabalin by Straube, et al., the authors found a significant improvement in the mental component score for both the 450 mg (P<0.01) and the 600 mg (P<0.05) doses of pregabalin.74 Combining the data for all 3 doses may explain the difference in our results. It is important to recognize that although significant, the absolute mean differences noted between the active drug and placebo ranged between 1.0 to 1.5 on a scale of 0–100, which is small and questions the clinical significance of this change. This may explain why the significant difference noted with milnacipran and duloxetine compared to placebo did not translate into any difference between the drugs in the meta-analysis. Our results differed from the analysis by Arnold on the 4 placebo-controlled trials of duloxetine which found a statistically significant difference on the physical component score as well (1.08; 95% CI, 0.12 to 2.03; P<0.05). In reviewing the Arnold, et al. pooled analysis of the 4 placebo-controlled duloxetine trials,65–67, 69 it appeared that the authors used unpublished individual patient data which was not provided in their report. This study was given a poor quality rating because of its risk of selection bias due to the failure to perform and report on a systematic search process, not reporting trial details or outcome data of individual trials, and failure to grade the quality of their results.57 No trial of amitriptyline reported on this outcome.

In summary, there was low to insufficient evidence that all drugs are superior to placebo on the total score of the Fibromyalgia Impact Questionnaire with no difference between drugs. Milnacipran was found to have a small but significant improvement on the Medical Outcomes Study 36-item Short-Form Health Survey physical and mental component compared with placebo and duloxetine was found to have a small but significant improvement on the mental component summary. No difference was found between duloxetine, milnacipran, or pregabalin on this measure.

Other outcomes

Given the significant variability and sparsity of reporting data on outcomes of sleep disturbance, health-related quality of life, and depressed mood in the amitriptyline trials, we did not analyze these outcomes. Hauser, et al. performed a meta-analysis of placebo-controlled trials of duloxetine, milnacipran, and pregabalin, and did perform a pooled analysis and comparative meta-analysis for these outcomes.49 Since their report was released, 2 additional milnacipran trials have been published,52, 53 with only 1 providing additional evidence on the outcome of sleep.53 These new results are consistent with prior studies. Hauser, et al. found that duloxetine was significantly superior to milnacipran on all 3 outcomes but was superior to pregabalin only on the outcome of improvement in depressed mood (Table 5). They found that pregabalin was superior to milnacipran on improvement in sleep disturbance and health-related quality of life, whereas milnacipran was superior to pregabalin on improvement in depressed mood (Table 5).

Table 5. Indirect analysis of placebo-controlled trials of pregabalin, milnacipran, and duloxetine for fibromyalgia.

Table 5

Indirect analysis of placebo-controlled trials of pregabalin, milnacipran, and duloxetine for fibromyalgia.

Comparisons to placebo
Gabapentin

One randomized, placebo-controlled, 12-week, fair-quality trial of 150 patients found that gabapentin 1800 mg (median) significantly improved pain severity, overall impact of fibromyalgia, global status, and sleep, but not tender point pain threshold, depression, or overall quality of life.78 Diagnosis of fibromyalgia was based on the 1990 American College of Rheumatology criteria. Patients were 90% female, 97% white, and had a mean age of 48 years. Duration of fibromyalgia was not reported. A total of 19% of patients had a current major depressive disorder and 9% had a current anxiety disorder.

For pain, gabapentin was superior to placebo in reducing average pain severity score (−44% compared with − 33%; P=0.015) and interference score (−53% compared with − 32%; P=0.032) from the Brief Pain Inventory and in the proportion of patients who achieved a 30% response rate (51% compared with 31%; P= 0.014; relative risk 1.65; 95% CI, 1.11 to 2.50; number needed to treat, 5), but not in improvement of the mean tender point pain threshold (+0.2 kg/cm2 compared with +0.1 kg/cm2; P=0.11). For overall impact fibromyalgia, there was a significantly greater reduction in Fibromyalgia Impact Questionnaire mean total score for gabapentin than placebo (− 43% compared with − 22%; P=0.001). For global status, there was a greater reduction Clinical Global Impression of Severity Scale scores for gabapentin than for placebo (− 29% compared with − 15%; P=0.002). Gabapentin also significantly reduced the Medical Outcomes Study Sleep Problems Index score (−40% compared with − 14%; P= 0.001). The difference between gabapentin and placebo did not reach statistical significance for increasing mean tender point threshold (+11% compared with +6%), reducing depression based on the Montgomery Asberg Depression Rating Scale score (−43% compared with − 19%; P=0.067) or improving quality of life based on the Medical Outcomes Study, Short-Form 36 (data not reported) for sleep, Montgomery Asberg Depression Rating Scale for depression, and Medical Outcomes Study 36-item Short-Form Health Survey.

Cyclobenzaprine

Compared with placebo, a significant reduction in pain severity was only found with cyclobenzaprine in the largest (N=120)38 of 3 fair-quality trials.38, 40, 41 Data on pain could not be pooled across trials due to heterogeneity in outcome assessment. The trials included a total of 172 primarily female patients (range, 83% to 100%) with mean ages ranging from 43 years41 to 49 years.38 None of the trials used the 1990 American College of Rheumatology criteria for diagnosing fibromyalgia. In 2 cases, this was because the trials were conducted prior to 1990.38, 40 History of fibromyalgia ranged widely in the 2 trials that reported this information.38, 40 One trial enrolled newly diagnosed patients(4.5 months),38 whereas mean duration of pain was 11.4 years in the second trial.40 In all 3 trials, cyclobenzaprine dosage was based on a flexible regimen, starting at 10 mg and going up to a maximum dosage of 40 mg per day (means not reported).

The only trial that found a significant reduction in pain severity for cyclobenzaprine involved 120 patients with newly diagnosed fibrositis; 44% of whom had primary fibrositis and 56% had fibrositis considered to be associated with trauma or arthritis.38 Patients were enrolled both from a university rheumatology clinic in Portland, Oregon and the Center for Arthritis and Back Pain in Philadelphia, Pennsylvania. Based on patient self-assessments (visual analog scale of 0–10), there was a significantly greater improvement with cyclobenzaprine than with placebo in pain (28% compared with 17%; P<0.02) and sleep (34% compared with 18%; P<0.02) at 12 weeks. Also, physicians’ evaluation of overall response to therapy found a significantly greater proportion of patients in the cyclobenzaprine group to have marked or moderate global improvement (34% compared with 16%; P<0.012; relative risk, 2.18; 95% CI, 1.12 to 4.36; number needed to treat, 5). However, at 12 weeks, there were no significant differences between cyclobenzaprine and placebo in duration of stiffness or fatigue, tender points, or physician-assessed muscle tightness or global pain.

In the smaller trials, there were trends favoring cyclobenzaprine over placebo in the number of patients with moderate to marked improvement in muscle pain at 6 weeks (43% compared with 18%; P not reported; N=40)40 and in change in pain severity at 4 weeks (unspecified pain scale, scores ranged 0 to 60) in the evening (− 8% compared with +4%; P not reported; N=12),41 but not in the morning (−12% for both). These nonsignificant findings may have been the result of limitations in statistical power due to the small sample sizes in these studies.

Selective serotonin reuptake inhibitors

Among placebo-controlled trials of citalopram (N=82),79, 80 fluoxetine (N=102),81, 82 and immediate-release and controlled-release paroxetine (N=40),83, 84 only fluoxetine, at a higher dose (45 mg), resulted in significantly greater improvements than placebo in pain, fatigue, and Fibromyalgia Impact Questionnaire Total Score.81

These trials enrolled 94% to 100% female patients with mean ages ranging from 31 years83 to 50 years.82 All but 1 trial82 used the 1990 American College of Rheumatology criteria for diagnosing fibromyalgia. History of fibromyalgia ranged from 10 years80 to 13 years.82 Dosages ranged from 20 mg to 45 mg across all trials. All but 1 trial was rated fair quality. The remaining trial was rated poor quality due to concern over the potential confounding effects of higher numbers of high school graduates (90.5% compared with 61.9%; P=0.03) and longer disease duration (16.1 years compared with 9.6 years; P=0.05) in the fluoxetine group at baseline and the impact of excluding 43% of patients from the analyses.82

Due to small sample sizes and incomplete reporting of outcome data, evidence from these trials was insufficient for indirect meta-analysis and did not permit conclusions about comparative effectiveness. Table 6 summarizes the results for each selective serotonin reuptake inhibitor compared with placebo for the most consistently reported outcomes.

Table 6. Selective serotonin reuptake inhibitor compared with placebo: Mean changes in symptom severity.

Table 6

Selective serotonin reuptake inhibitor compared with placebo: Mean changes in symptom severity.

Among selective serotonin reuptake inhibitors, only fluoxetine, at a higher dose (45 mg), resulted in significantly greater improvements than placebo in pain, fatigue, and Fibromyalgia Impact Questionnaire Total Score in 1 small trial of 51 patients (Table 6).81 Fluoxetine also resulted in a significantly greater reduction than placebo on the Fibromyalgia Impact Questionnaire depression subscore (−33% compared with +44%; P=0.01). Effects on sleep and global status were not reported.

As for paroxetine, although the controlled-release formulation (39.1 mg) did not result in a significant reduction in the Fibromyalgia Impact Questionnaire pain subscale (Table 6) and did not significantly increase the number of patients with a 50% reduction in Fibromyalgia Impact Questionnaire Total Score (26% compared with 14%; P=0.08), it was superior to placebo on the primary outcome of 25% reduction in Fibromyalgia Impact Questionnaire Total Score and on a few secondary outcome measures.84 Controlled-release paroxetine was superior to placebo in reducing the Fibromyalgia Impact Questionnaire Total Score (Table 6), the Fatigue Subscore (Table 6), and in improving global status (Clinical Global Impression-Improvement, P<0.005), but not for reducing depression (Beck Depression Inventory) or disability (Sheehan Disability Scale).84 Evidence from the second, single-blind placebo-controlled trial of immediate-release paroxetine 20 mg (N=40) did not contribute useful information as the analyses focused on evaluating change from baseline in each group, respectively, and did not directly compare immediate-release paroxetine to placebo.83

Finally, citalopram demonstrated few advantages over placebo in only 1 (N=40)79 of 2 studies.79, 80 Neither trial found a significant improvement with citalopram for pa in or fatigue (Table 6). Citalopram was superior to placebo in reducing the total score on the Montgomery Asberg Depression Rating Scale (P<0.01, data not reported),79 but not in reducing the Beck Depression Scale score.80 For sleep, scores on the Montgomery Asberg Depression Rating Scale sleep item improved significantly in the citalopram group, but not in the placebo group in 1 trial,79 and there was no difference in Visual Analogue Scale score between the citalopram and placebo groups in the second trial.80 There was no significant difference between citalopram and placebo in global status in either trial.

Key Question 2. For adults with fibromyalgia, what are the comparative harms of included interventions?
2a. When used as monotherapy?
2b. When used as adjunctive therapy?

Summary of Findings

Direct evidence

  • Although there were some significant differences between drugs in overall adverse events, they did not produce any differences in withdrawals due to adverse events
    • Overall and anticholinergic-type adverse events were significantly more frequent with amitriptyline 100 mg than with immediate-release paroxetine 20 mg, but withdrawals due to adverse events did not differ
    • Overall adverse events were significantly greater for nortriptyline than for amitriptyline, but individual adverse events and withdrawals due to adverse events were not significantly different
    • No significant differences were found between cyclobenzaprine and amitriptyline in any harms outcomes.

Indirect evidence

Indirect meta-analysis
Comparisons to placebo
  • Gabapentin: Despite greater rates of various individual adverse events (e.g., dizziness, sedation, light headedness, and weight gain), withdrawals due to adverse events were not significantly greater with gabapentin than placebo
  • Cyclobenzaprine: Dry mouth, drowsiness, and overall adverse events were more frequent with cyclobenzaprine than placebo, but withdrawals due to adverse events did not differ between treatment groups
  • Selective serotonin reuptake inhibitors: Compared with placebo, significant increases in adverse events were only found in the largest trial of controlled-release paroxetine, including drowsiness, dry mouth, and ejaculatory problems.

Detailed Assessment

Direct evidence

Direct evidence regarding the comparative harms among included interventions was limited to 3 fair-quality randomized controlled trials that evaluated amitriptyline with cyclobenzaprine,43 nortriptyline,45 and immediate-release paroxetine.46 A detailed description of the trial design and patient characteristics can be found above in Key Question 1.

Although there were some significant differences between drugs in overall adverse events, they did not produce any differences in withdrawals due to adverse events (Table 7).43, 45, 46 Among individual adverse events, the only significant difference came from the trial that compared immediate-release paroxetine 20 mg and amitriptyline 100 mg, in which anticholinergic-type adverse events (e.g., dry mouth, constipation, urinary retention) were more frequent with such a high dose of amitriptyline (41% compared with 9%; P<0.004).46 Otherwise, there were no significant differences between amitriptyline and either cyclobenzaprine43 or nortriptyline.45 In the trial that compared amitriptyline to cyclobenzaprine, dry mouth, somnolence, dizziness, and weight gain were cited as being the most frequently reported adverse events, but incidence rates were not reported.43 The most frequent adverse events for amitriptyline and nortriptyline were dry mouth (10% compared with 16%), abdominal pain (10% compared with 18%), and dizziness (5% compared with 10%).45

Table 7. Adverse events in head-to-head trials.

Table 7

Adverse events in head-to-head trials.

Indirect evidence

We found 3 good-quality systematic reviews that assessed multiple drugs for the treatment of fibromyalgia and analyzed harm.48, 50, 85 One of these reviews performed an indirect meta-analysis to compare the effectiveness and harms between duloxetine, milnacipran, and pregabalin in fibromyalgia from placebo-controlled trials of individual drugs.85 Although it was a recent review, we chose to update the indirect meta-analysis to include evidence from 2 new trials for milnacipran that are now available.52, 53 We have compared our results to the Hauser, et al. analysis and discussed where applicable below. One systematic review of amitriptyline reported on 10 randomized controlled trials but did not perform a meta -analysis due to large clinical variability and statistical heterogeneity.50 The results are considered as appropriate below. One systematic review pooled data of pregabalin and gabapentin and reported on a class effect size rather than comparative data, which limited its usefulness for this report.48

Of the 8 drugs included in this review, we found that only 4 drugs had multiple trials with sample sizes adequate to perform a comparative analysis. All trials used the drugs as monotherapy and no trial evaluated the drugs as adjunctive therapy. We performed a meta-analysis of 6 placebo-controlled trials of amitriptyline,16, 39, 43, 45, 54, 55 4 placebo-controlled trials of pregabalin,57–60 5 placebo -controlled trials of milnacipran,52, 53, 61–64 and 4 placebo-controlled trials of duloxetine 65–69 to determine comparative harm. Given that Hauser, et al. performed a recent meta-analysis to determine comparative harm, and given that the harms data from the 2 additional milnacipran trials were consistent with prior results, we elected to study the differences between the drugs on measures of overall withdrawal, overall adverse events, and withdrawal due to adverse events.49, 52, 53 None were of long duration (8 to 28 weeks) with the amitriptyline trials being of the shortest duration (mean 9 weeks). Sample size was small for all of the amitriptyline trials (N=22 to 126) but generally moderate to large for the rest (N=125 to 1025). All of the trials were rated fair quality. One poor-quality trial of amitriptyline was excluded from our analysis.73 The trials of duloxetine, milnacipran, and pregabalin were industry sponsored whereas only 1 of the amitriptyline trials were industry sponsored.16 The baseline demographics for all of the trials were similar, with the majority of patients being middle aged (mean age range 40–53 years), white (65% to 93%), and female (82% to 100%). For amitriptyline, we excluded data on the lowest dose of 12.5 mg once daily from 1 trial, and for pregabalin we excluded data on the lowest dose of 150 mg once daily. Otherwise we combined data on different daily dosage groups including amitriptyline 25–50 mg, pregabalin 300–600 mg, milnacipran 100–200 mg, and duloxetine 40–120 mg. We reported the outcomes of clinical importance and with adequate data to perform the analysis. We performed our analysis using short-term data (8–15 weeks) as this was consistent across most studies. We excluded trials shorter than 8 weeks from our analysis.42, 44 We performed a sensitivity analysis to determine if differences in outcome were seen using both short- and long-term data. The results and quality of these trials are summarized in Evidence Tables 1 and 2.

Overall withdrawal

Pooled analysis of short-duration (8–15 weeks) placebo-controlled trials found no difference in overall withdrawal of amitriptyline, pregabalin, and duloxetine compared with placebo whereas milnacipran was found to have a small but significant increase in overall withdrawal compared with placebo (Table 8). There was low evidence that no differences exist between the drugs on overall withdrawal when we included all of the trials (Table 9). This result held true for our sensitivity analysis using all trials (6–28 weeks) of pregabalin, duloxetine, and milnacipran. There was insufficient long-term data to include amitriptyline in this analysis. When we excluded the Carette 1986 trial that was conducted prior to the development of the 1990 American College of Rheumatology criteria for fibromyalgia, amitriptyline appeared to have less overall withdrawal, however it is felt that this may be reflective of the lower sample size (N =208).

Table 8. Pooled effectiveness of amitriptyline, pregabalin, milnacipran, and duloxetine compared with placebo (8–15 weeks).

Table 8

Pooled effectiveness of amitriptyline, pregabalin, milnacipran, and duloxetine compared with placebo (8–15 weeks).

Table 9. Indirect analysis of placebo-controlled trials in fibromyalgia.

Table 9

Indirect analysis of placebo-controlled trials in fibromyalgia.

Overall adverse events

Pooled analysis of placebo-controlled trials of amitriptyline, pregabalin, duloxetine, and milnacipran found a significant increase in overall adverse events of pregabalin, duloxetine, and milnacipran compared with placebo, whereas amitriptyline was found to be no different than placebo on this outcome (Table 8). This held true when we excluded the Ginsberg trial, which used a sustained-release formulation of amitriptyline, and the Carette 1986 trial.39, 55 Given the small sample size in the pooled analysis of amitriptyline, we felt that the data was insufficient to make any conclusive statements. There was low evidence that no differences exist between the other drugs on overall adverse events (Table 9). This result held true for our sensitivity analysis using all trials (6–28 weeks) of pregabalin, duloxetine, and milnacipran. There was insufficient long-term data to include amitriptyline in this analysis.

Withdrawal due to adverse events

Pooled analysis of placebo-controlled trials of amitriptyline, pregabalin, duloxetine, and milnacipran found a significant increase in withdrawal due to adverse events of pregabalin, duloxetine, and milnacipran compared with placebo whereas amitriptyline was found to be no different than placebo on this outcome (Table 8). Given the small sample size in the pooled analysis of amitriptyline, we felt that the data was insufficient to make any conclusive statements regarding this outcome. There was low evidence that no differences exist between the other drugs on withdrawal due to adverse events (Table 9). This result held true for our sensitivity analysis using all trials (6–28 weeks) of pregabalin, duloxetine, and milnacipran. There was insufficient long-term data to include amitriptyline in this analysis.

Other adverse events

Hauser, et al. performed a good-quality systematic review with an indirect meta-analysis to compare the harms between duloxetine, milnacipran, and pregabalin in fibromyalgia from placebo-controlled trials of individual drugs.49 No drug-related deaths were reported and all drugs were generally well tolerated. They found that duloxetine and milnacipran had significantly greater reporting of headache and nausea compared with pregabalin but no difference between each other (Table 10). Duloxetine also had increased diarrhea compared with milnacipran and pregabalin and no difference between the later 2 drugs (Table 10). All drugs had a significant increase in dry mouth, constipation, and dizziness compared with placebo and no difference between the drugs.49 Both duloxetine and pregabalin had a significant increase in fatigue and somnolence compared with placebo and no difference between the 2 drugs.49

Table 10. Indirect analysis of harms from placebo-controlled trials of pregabalin, milnacipran, and duloxetine for fibromyalgia.

Table 10

Indirect analysis of harms from placebo-controlled trials of pregabalin, milnacipran, and duloxetine for fibromyalgia.

Reporting of hyperhidrosis was unique to duloxetine and milnacipran with no differences between the drugs on this outcome (relative risk, 1.14;95% CI, − 5.08 to 7.36).49

Milnacipran was the only drug that reported tachycardia (number needed to harm, 20.6; 95%CI, 15.1 to 29.1).49 Pregabalin was the only drug that reported weight gain and peripheral edema, both of which were significantly greater than placebo (relative risk, 4.58;95% CI, 2.44 to 6.82; and relative risk, 3.52;95% CI, 2.01 to 6.18 respectively).49 All drugs reported rare serious adverse events including risk of suicide (duloxetine 1.1% based on one 1-year trial, milnacipran 1.3% in those with depression at baseline, and pregabalin <1%) and hepatic-related adverse events.49

Nishishinya, et al. conducted a systematic review of placebo-controlled trials of amitriptyline but did not perform a pooled analysis. Of 6 trials reporting adverse events, they found that the mean adverse event rate for amitriptyline was 51.84% (2.8% to 95%) compared with 36.63% (2.8 % to 80%) for placebo.50 The high placebo event rate questioned the validity of the results given that 2 trials had higher adverse event rates in the placebo arm compared with the amitriptyline arm.50 Amitriptyline was generally well tolerated in all the trials with no severe or life threatening events reported.50 Somnolence, dry mouth, gastrointestinal symptoms, and weight gain were the most frequently reported adverse events and there were no differences in withdrawal due to adverse events compared with placebo.50

Comparisons to placebo
Gabapentin

Over 12 weeks, dizziness (25% compared with 9%; P< 0.05; relative risk, 2.86;95% CI, 1.3 to 6.29, number needed to harm, 6), sedation (24% compared with 4%; P<0.001; relative risk, 6.33; 95% CI, 2.12 to 19.54; number needed to harm, 5), light headedness (15% compared with 1%; P<0.01; relative risk, 11.00; 95% CI, 1.91 to 65.55; number needed to harm, 7), and weight gain (8% compared with 0; P< 0.05; relative risk, 13.00; 95% CI, 1.61 to infinity; number needed to harm, 12) were more frequent with gabapentin 1800 mg than placebo.78 But, withdrawals due to adverse events did not differ significantly in the gabapentin and placebo groups (16% compared with 9%; P=0.34).

Cyclobenzaprine

Data on harms were inconsistently reported across placebo-controlled trials of cyclobenzaprine. Although the incidence of dry mouth (pooled rates, 56% compared with 20%; pooled relative risk, 2.75; 95% CI, 1.71 to 4.43; number needed to harm, 3),38, 40 drowsiness (34% compared with 17%; P<0.01; relative risk, 1.96; 95% CI, 1.04 to 3.82; number needed to harm, 6),38 and overall adverse events (89% compared with 64%; relative risk, 1.39; 95% CI, 1.14 to 1.76; number needed to harm, 4)38 was greater with cyclobenzaprine than placebo, withdrawals due to adverse events did not differ between treatment groups (pooled rates: 6% compared with 2%; pooled relative risk, 2.56; 95% CI, 0.48 to 13.59).38, 40, 41

Selective serotonin reuptake inhibitors

Adverse events were sparsely reported in placebo-controlled trials of selective serotonin reuptake inhibitors.79–84 Compared with placebo, withdrawals due to adverse events were similar for citalopram (14% compared with 0%; P not reported; N=40)79 and for controlled-release paroxetine (7% compared with 2%; P not reported; N=116),84 but were not reported for fluoxetine. Overall adverse events were only reported in 1 trial of controlled-release paroxetine, and were not significantly different from placebo (65% compared with 59%; P not reported).84 Compared with placebo, in the largest trial (N=116), controlled-release paroxetine was the only selective serotonin reuptake inhibitor to significantly increase any specific adverse events, including drowsiness (26% compared with 7%; P not reported), dry mouth (36% compared with 9%; P not reported), and ejaculatory problems (66% compared with 2%; P not reported).84

Key Question 3. Are there subgroups of patients based on demographics (age, racial or ethnic groups, and gender), socioeconomic status, other medications, or comorbidities for which any included drugs are more effective or associated with fewer harms?

Summary of Findings

Direct evidence

Indirect evidence

  • Nine individual trials performed subgroup analysis
  • The majority of patients in all individual trials were middle-aged, white (84% to 91%), and female (89% to 100%), and there was a high prevalence of baseline anxiety and depression
  • Duloxetine was no different than placebo in pain response in male patients (−1.24; standard error, 0.4 compared with −1.25; standard error, 0.5; P=0.969), those ≥ 65 years, (−1.92; standard error, 0.3 compared with −1.50; standard error, 0.4; P=0.374), and nonwhite patients (−1.70; standard error, 0.3 compared with −1.37; standard error, 0.3, P=0.386); this result should be interpreted with caution given the small number of patients in each group
  • Duloxetine: Based on 2 trials, effect on pain was estimated to be 61% to 86% after accounting for the indirect effect on depression
  • Pregabalin: Based on 1 trial, effect on pain was estimated to be 75% after accounting for the indirect effect on anxiety and depression
  • Milnacipran: Greater improvement in pain reduction was seen in nondepressed patients compared with depressed patients; however, the small number of patients in each group limits the interpretation of these results
  • Cyclobenzaprine: Race, age, nor sex were found to influence effectiveness as compared to placebo
  • Selective serotonin reuptake inhibitors: History of depression was not associated with response to either fluoxetine or controlled-release paroxetine, nor was anxiety associated with response to controlled-release paroxetine.

Detailed Assessment

We identified 10 trials that performed subgroup analysis of the included drugs.38, 43, 62, 65, 66, 78, 81, 87-89 The baseline demographics for all of the trials were similar, with the majority of patients being middle aged (mean age range 44–50 years), white (84% to 91%), and female (89% to 100%). Analysis of comparative effectiveness in men, nonwhites, and older patients was limited by the small number of patients in these subgroups and the corresponding lack of reporting of data in the individual trials. In the 3 trials that reported time since diagnosis of fibromyalgia, the duration ranged from 49 months to 129 months.43, 62, 87

Direct evidence

Only 1 head-to-head trial addressed drugs used for fibromyalgia in subgroups of the population, and only assessed the impact of age.43 In a fair-quality randomized controlled trial of 208 patients who met the 1990 American College of Rheumatology criteria for fibromyalgia, variation in age did not differentiate response at 6 months to either amitriptyline 50 mg or cyclobenzaprine 30 mg. Response was defined as meeting at least 4 of the following 6 criteria: 50% improvement in pain, sleep, fatigue, patient global assessment, or physician global assessment, and increase of 1 kg in mean total myalgic score.

Indirect evidence

The small number of patients in each subgroup and the limited reporting of subgroup data did not allow for indirect meta-analysis of the outcome measures for comparative effectiveness.

Duloxetine

One randomized double-blind fair-quality placebo-controlled trial restricted inclusion of patients to females only (N=354).66 The results for all outcome measures did not differ from the indirect meta-analysis revealing that duloxetine either 60 mg once daily or 60 mg twice daily is more effective than placebo at 12 weeks on outcomes of pain, Fibromyalgia Impact Questionnaire, and patient global impression of improvement. The 50% response rate in pain was significant and ranged from 41% for duloxetine compared with 23% for placebo (P=0.003). A pooled subgroup analysis of female patients in all 4 duloxetine placebo-controlled trials(n =1262) found a statistically significant difference in pain reduction compared to placebo with mean least squares change in 24-hour pain measures on the brief pain inventory, a 10-point scale of − 1.74 (standard error, 0.1) for the duloxetine group and − 1.10 (standard error, 0.1) for the placebo group (P<0.001).65–69

A pooled analysis of all 4 duloxetine placebo-controlled trials evaluated the response of male patients, those age 65 or older, and nonwhite patients.75 Unlike the results of the total population, the difference in pain response between duloxetine and placebo was not significant in male patients, n=70, (−1.24; standard error, 0.4 compared with −1.25; standard error, 0.5; P=0.969), those ≥ 65 years, number of patients not reported, (− 1.92; standard error,0.3; compared with − 1.50; standard error, 0.4; P=0.374), and nonwhite patients, n=147, (−1.70; standard error, 0.3 compared with − 1.37; standard error, 0.3; P=0.386). Despite the pooling of 4 moderate sized trials, there remained a small number of patients in each subgroup resulting in the analysis being underpowered to detect a true difference if a difference exists. Further analysis demonstrated that treatment by subgroup interaction was not significant in their overall analysis of results on pain measures (sex, P=0.320; age, P=0.362; race, P=0.180).

Two trials of duloxetine in women performed an analysis to determine whether the treatment effect of duloxetine in fibromyalgia was independent of its effect on major depressive disorder or anxiety.65, 66 A regression model was used to determine if the treatment has an effect on pain reduction after accounting for the indirect effect through improvement of depressive symptoms. In Arnold 2004, the direct effect of duloxetine on reduction of pain accounted for 61.1% to 83.3% of the total treatment effect, with an indirect effect through improvement in depressive symptoms accounting for 15.3% to 38.5%, and improvement in anxiety accounting for 0.5% to 1.5% of the total effect. 65 In Arnold 2005, the direct effect of duloxetine on reduction of pain accounted for 75.6%to 85.9% of the total treatment effect with an indirect effect through improvement in depressive symptoms accounting for 13.1% to 24.4%. In a subgroup of 91 patients with a current diagnosis of major depressive disorder, a significant reduction in the Brief Pain Inventory average pain severity score was demonstrated between duloxetine and placebo (60 mg once daily, P=0.005 and 60 mg twice daily, P=0.003).

Milnacipran

One 12-week randomized placebo-controlled fair-quality trial evaluated the effect of depression comorbidity on response to once or twice-daily milnacipran in patients diagnosed with fibromyalgia.62, 64 The mean dose was 174 mg for the once daily group and 191 mg for the twice daily group. There was statistically less baseline depression in the once daily milnacipran group (n=3; 7%) compared with the milnacipran twice daily group (n=8; 16%) and placebo group (n=9; 32%). Greater improvement in pain reduction was seen in nondepressed patients compared with depressed patients treated with milnacipran; however, the small number of patients in each group and the high placebo response rate in depressed patients limited the interpretation of these results. The authors attributed this difference to the higher placebo response rate among depressed patients. For twice daily milnacipran, 25% of depressed patients and 37% of nondepressed patients had 50% improvement in daily pain score compared with 44% and 0% respectively for placebo.

Pregabalin

One 8-week randomized double-blind placebo-controlled trial (N=529)compared pregabalin 450 mg once daily to placebo and evaluated the impact of anxiety and/or depression on the effectiveness of pregabalin in reducing pain compared with placebo. 87 At baseline patients completed the Hospital Anxiety and Depression Scale which consists of two 7-item subscales, 1 for depression and 1 for anxiety. A regression model was used to determine if the treatment has an effect on pain reduction after accounting for the indirect effect through improvement of depressive symptoms. Baseline anxiety and depression fell in the mild range for both, with 29% of patients being free of anxiety and 44% of patients being free of depression at baseline indicating significantly higher baseline anxiety (P<0.0001). Baseline mean pain was 7.0 (standard deviation, 1.3). There was a statistically significant reduction in pain in the pregabalin group with a mean score of 4.94 compared with 5.88 in the placebo group (P=0.001). Using a path analysis based on linear regression models, they estimated that improvement in anxiety and depression each indirectly attributed 12.4% of pain relief, with 75.3% attributable to the direct treatment effect of pregabalin. This trial excluded patients with clinically significant psychiatric illness at the time of enrollment which may have limited the interpretation of these results. A pooled post-hoc analysis of the 4 placebo-controlled pregabalin trials was performed by Emir, et al. to evaluated whether clinical characteristics of patients at baseline influenced the magnitude of pain response to pregabalin.89 They used a covariate inter action to estimate mean pain changes and least square means across different levels of baseline covariates for 2061 patients, including baseline anxiety and depression. Of 2032 patients with this data, they found no significant interactions between treatment effectiveness and baseline anxiety or depressive symptoms (anxiety, P=0.654; depression, P=0.689).89 They did find that patients with a higher pain score at baseline, a higher sleep score based on the medical outcomes sleep score index (MOS, 0–100 scale), and older age were all associated with a more robust and significantly greater improvement in pain (P<0.0001 for each).89

Cyclobenzaprine

In one 12-week trial of 120 patients, race, age, nor sex were found to influence the effectiveness of cyclobenzaprine compared with placebo. 38

Selective serotonin reuptake inhibitors

Two fair-quality randomized controlled trials consistently found that a history of depression was not associated with response to either fluoxetine81 or controlled-release paroxetine.88 In one 12-week trial of 60 patients, analysis of covariance revealed no significant interaction between the effect of fluoxetine 45 mg on pain and history of major depressive disorder, baseline level of depression, or improvement in depression.81 In a post -hoc logistic regression analysis of data from a 12-week placebo-controlled trial of controlled-release paroxetine in 116 patients with fibromyalgia, history of depression and/or anxiety did not predict treatment response as measured by at least a 25% reduction in Fibromyalgia Impact Questionnaire score (odds ratio, 0.66; 95% CI, 0.29 to 1.49).88

Footnotes

a

The Drug Effectiveness Review Project uses a modified PRISMA flow diagram.1

Copyright © 2011, Oregon Health & Science University.
Cover of Drug Class Review: Drugs for Fibromyalgia
Drug Class Review: Drugs for Fibromyalgia: Final Original Report [Internet].
Smith B, Peterson K, Fu R, et al.
Portland (OR): Oregon Health & Science University; 2011 Apr.

Download

PubMed Health Blog...

read all...

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...