Treatment of Fibromyalgia Syndrome with Antidepressants a Meta-Analysis

CLINICAL REVIEW CLINICIAN’S CORNER

Treatment of Fibromyalgia Syndrome With Antidepressants A Meta-analysis

Winfried Häuser, MD Context Fibromyalgia syndrome (FMS) is a chronic pain disorder associated with mul- Kathrin Bernardy, PhD tiple debilitating symptoms and high disease-related costs. Effective treatment op- Nurcan Üc¸eyler, MD tions are needed. Claudia Sommer, MD Objectives To determine the efficacy of antidepressants in the treatment of FMS by performing a meta-analysis of randomized controlled clinical trials. IBROMYALGIA SYNDROME (FMS) Data Sources MEDLINE, PsycINFO, Scopus, and the Cochrane Library databases has an estimated prevalence in were searched through August 2008. Reference sections of original studies, meta- North America and Europe of analyses, and reviews on antidepressants in FMS were reviewed. 0.5% to 5.8%.1 According to the Study Selection Randomized placebo-controlled trials with tricyclic and tetracyclic criteriaF of the American College of antidepressants (TCAs), selective serotonin reuptake inhibitors (SSRIs), serotonin and Rheumatology (ACR), FMS is defined noradrenaline reuptake inhibitors (SNRIs), and monoamine oxidase inhibitors (MAOIs) as chronic widespread pain and ten- were analyzed. derness at a minimum of 11 of 18 de- Data Extraction and Data Synthesis Two authors independently extracted data. 2 fined tender points. Other symptoms Effects were summarized using standardized mean differences (SMDs) by a random- of FMS are fatigue and nonrestorative effects model. sleep. Most patients report additional Results Eighteen randomized controlled trials (median duration, 8 weeks; range, 4-28 somatic and psychological symp- weeks) involving 1427 participants were included. Overall, there was strong evidence 3,4 toms. Patients with FMS experience for an association of antidepressants with reduction in pain (SMD, −0.43; 95% con- disability and reduced health-related fidence interval [CI], −0.55 to −0.30), fatigue (SMD, −0.13; 95% CI, −0.26 to −0.01), quality of life (HRQOL).5 Fibromyal- depressed mood (SMD, −0.26; 95% CI, −0.39 to −0.12), and sleep disturbances (SMD, gia syndrome is also associated with −0.32; 95% CI, −0.46 to −0.18). There was strong evidence for an association of an- high direct6,7 and indirect disease- tidepressants with improved health-related quality of life (SMD, −0.31; 95% CI, −0.42 related costs.8 Effective treatment of to −0.20). Effect sizes for pain reduction were large for TCAs (SMD, −1.64; 95% CI, FMS is therefore necessary for medi- −2.57 to −0.71), medium for MAOIs (SMD, −0.54; 95% CI, −1.02 to −0.07), and 9 small for SSRIs (SMD, −0.39; 95% CI, −0.77 to −0.01) and SNRIs (SMD, −0.36; 95% cal and economic reasons. CI, −0.46 to −0.25). Whether FMS is a distinct disorder or a manifestation of another underly- Conclusion Antidepressant medications are associated with improvements in pain, depression, fatigue, sleep disturbances, and health-related quality of life in patients ing disorder is controversial. The spec- with FMS. trum of possible underlying disorders JAMA. 2009;301(2):198-209 www.jama.com ranges from inflammatory arthritic diseases to depression. Others classify FMS Author Affiliations: Department of Internal Medi- Corresponding Author: Winfried Häuser, MD, Klini- 10 cine, Klinikum Saarbrücken, Saarbrücken, Germany (Dr kum Saarbrücken, Winterberg 1, D-66119 Saar- as a functional somatic syndrome. Häuser); Department of Anesthesiology, Emergency brücken, Germany (whaeuser@klinikum-saarbruecken Medicine and Pain Therapy, University of Saarland, .de). Saarbrücken (Dr Bernardy); Department of Psycho- Clinical Review Section Editor: Mary McGrae CME available online at somatic Medicine, MediClin Bliestal Clinics, Blieskas- McDermott, MD, Contributing Editor. We encour- www.jamaarchivescme.com tel, Germany (Dr Bernardy); and Department of Neu- age authors to submit papers for consideration as a and questions on p 227. rology, University of Würzburg, Würzburg, Germany Clinical Review. Please contact Mary McGrae (Drs Üc¸eyler and Sommer). McDermott, MD, at [email protected].

Downloaded From: https://jamanetwork.com/ on 09/28/2021 ANTIDEPRESSANT THERAPY IN FIBROMYALGIA SYNDROME

Evidence-based guidelines on the dence-based guidelines11,12 were screened an additional outcome. When research- management of FMS from the Ameri- manually and independently by 2 of us ers reported more than 1 measure for can Pain Society11 and the European (K.B., W.H.). an outcome, we used the following League Against Rheumatism12 in- priority for inclusion in the meta- cluded published literature through Study Selection analysis: 2004 and 2005, respectively. Antide- Study inclusion criteria were as fol- 1. Pain: visual analog scale (VAS), pressants are the drugs most often stud- lows: use of recognized criteria to VAS Fibromyalgia Impact Question- ied for treatment of FMS. However, no define FMS (ACR,2 Smythe and Mol- naire (FIQ), Numeric Rating Scale meta-analyses on antidepressant dofsky,18 or Yunus19); RCT design (NRS), other pain questionnaire therapy for FMS have been published with a control group receiving phar- 2. Fatigue: VAS, VAS FIQ, other since 2000.13,14 macological placebo; and treatment questionnaire We therefore performed a meta- with antidepressants (tricyclic and tet- 3. Sleep: VAS, VAS FIQ, NRS, other analysis with 3 goals: to evaluate the ef- racyclic antidepressants [TCAs], selec- questionnaire fects of treatment with antidepres- tive serotonin reuptake inhibitors 4. Depressed mood: VAS, VAS FIQ, sants on FMS-related symptoms; to [SSRIs], serotonin and noradrenaline other questionnaire examine possible differences in the ef- reuptake inhibitors [SNRIs], or mono- 5. HRQOL: FIQ total score, other ficacy of distinct antidepressant classes amine oxidase inhibitors [MAOIs]). HRQOL scale. in therapy for FMS; and to determine We did not include studies assess- The van Tulder test (11 items)21 and the internal validity (methodological ing cyclobenzaprine, S-adenosylme- the Jadad test (5 items)22 were applied quality) and external validity (gener- thionine, or combinations of antide- for assessing methodological quality. alizability) of randomized controlled pressants. Cyclobenzaprine combines The van Tulder items were used to ar- trials (RCTs) with antidepressants in characteristics of an antidepressant and bitrarily classify quality as high (scores FMS. a muscle relaxant. S-adenosylmethio- 8-11), moderate (scores 5-7), or low nine is a dietary supplement. We con- (scores 1-4). The Jadad score was used METHODS tacted corresponding authors of RCTs to classify quality as high (score 5), The meta-analysis was performed ac- with incomplete data presentation (eg, moderate (score 4), or low (scores 1-3). cording to the QUORUM guidelines missing means, standard deviations of (Quality of Reporting of Meta-analyses)15 pretest and posttest data, or standard Data Synthesis and Analysis and the recommendations of the Coch- deviations of change scores). Studies in We analyzed intention-to-treat data rane Collaboration.16 which only categorical data were pro- whenever available. For the compari- vided and those for which we were not son of proportions, the ␹2 test was ap- Data Sources and Searches able to obtain missing data were ex- plied. Nonparametric tests (Mann- The electronic databases screened were cluded. Whitney test, Kruskal-Wallis test) were MEDLINE (1966 through August 2008), used for comparing continuous vari- PsycINFO (1966 through August 2008), Data Extraction ables. A 2-sided P value of .05 or lower Scopus (1980 through August 2008), and Two of us (K.B., W.H.) independently was considered significant. Meta- the Cochrane Library (1993 through Au- screened the titles and abstracts of po- analyses were conducted using Rev- gust 2008). Using Medical Subject Head- tentially eligible studies identified. The Man analyses software (RevMan ings terms, searches were limited to hu- full text articles were examined inde- 4.2.10).23 man and performed for all languages. The pendently by 2 of us (C.S., W.H.) to de- Because most outcomes were pre- keyword fibromyalgia was used in com- termine whether they met the inclu- sented as continuous data (mean value bination with tricyclic antidepressant or sion criteria. Two of us (N.U., W.H.) or mean changes), we used either the serotonin reuptake inhibitors or mono- independently extracted data (study weighted mean differences (WMDs) or amine oxidase inhibitors or antidepres- characteristics and results) using data the standardized mean difference sant or antidepressive agents and random- extraction forms. Point estimates for se- (SMDs) as effect measures. Weighted ized controlled trial or controlled clinical lected variables were extracted and mean differences were calculated when trial or review. After consulting with the checked by the other 2 reviewers. We the outcome measure in all trials was German center of the Cochrane Collabo- used ␬ statistics to assess agreement be- determined on the same scale, SMDs ration, we did not use a filter such as the tween reviewers. All discrepancies were when outcomes were measured using highly sensitive search strategy17 be- rechecked and consensus was achieved different scales. To calculate WMDs or cause limiting the number of reports with by discussion. SMDs, we used means and change a filter did not seem reasonable consid- We selected the following outcome scores and their standard deviations. ering the potential number of studies. measures, which are features of FMS20: When only the standard error was re- Reference sections of relevant original ar- pain, fatigue, sleep, and depressed ported, it was converted into standard ticles, reviews, meta-analyses,13,14 and evi- mood. Health-related quality of life was deviation.24

Downloaded From: https://jamanetwork.com/ on 09/28/2021 ANTIDEPRESSANT THERAPY IN FIBROMYALGIA SYNDROME

Potential publication bias (ie, the as- tion of pain (SMD, −0.43; 95% confi- Figure 1. Study Selection sociation of publication probability with dence interval [CI], −0.55 to −0.30; Ͻ 337 Potentially relevant studies identified the statistical significance of study re- P .001), fatigue (SMD, −0.13; 95% CI, sults) was investigated using visual as- −0.26 to −0.01; P=.04), and depressed 303 Excluded sessment of the funnel plot (plots of mood (SMD, −0.26; 95% CI, −0.39 to 85 Not about antidepressants Ͻ 104 Duplicate publications effect estimates against sample size) cal- −0.12; P .001) and improved sleep 106 Reviews culated by RevMan Analyses software. (SMD, −0.32; 95% CI, −0.46 to −0.18; 8 No control group Publication bias may lead to asymmetri- PϽ.001) and HRQOL (SMD, −0.31; 26 Ͻ 34 Studies retrieved for more cal funnel plots. Furthermore, we 95% CI, −0.42 to −0.20; P .001). Based detailed evaluation tested the sensitivity of our results to on Cohen categories for evaluating the potential unpublished studies using a magnitude of effect sizes, the effect of 10 Excluded 1 No recognized FMS criteria file-drawer test for meta-analysis. This antidepressant therapy was negligible 1 N-of-1 trial 1 Duplicate publication test determines how many negative for fatigue and small for remaining 3 Lack of antidepressant-only group studies with an effect size of D=0.01 outcomes. 4 Lack of pharmacological placebo group would be needed to negate our find- TABLE 1 gives a comparison of the ings (fail-safe number). If the fail-safe effect sizes of each antidepressant class. 24 Potentially appropriate studies to number exceeds the file-drawer num- There was strong evidence for the ef- be included in the meta-analysis ber, the results of the meta-analysis can ficacy of the TCA amitriptyline in re-

6 Excluded (necessary data not provided) be regarded as robust against poten- ducing pain (SMD, −1.64; 95% CI, tial reporting bias.27-29 The file-drawer −2.57 to −0.71; PϽ.001), fatigue (SMD, 18 Studies included in meta-analysis number is calculated as 5kϩ10; where −1.12; 95% CI, −1.87 to −0.38; P=.003), k is the number of study groups in the and sleep disturbances (WMD, −1.84; FMS indicates fibromyalgia syndrome. meta-analysis. 95% CI −2.62 to −1.06; PϽ.001). Based on Cohen categories, these effect sizes I2 statistics were used to measure RESULTS were large. The effect size on de- heterogeneity of the RCTs. If the I2 Study Selection pressed mood (WMD, −0.60; 95% CI, value was less than 50%, a fixed- The literature search yielded 337 cita- −4.53 to 3.33; P=.76) was not signifi- effects meta-analysis was applied. If tions. Initially, 34 studies met our in- cant. The effect on HRQOL was small the I2 value was 50% or more, the clusion criteria. The excluded 303 stud- (WMD, −0.31; 95% CI, −0.60 to −0.01; random-effects meta-analysis was ies contained duplicate publications, P=.04). used.23 We used Cohen categories25 to review articles, uncontrolled studies, There was strong evidence for the ef- evaluate the magnitude of the effect and studies without an antidepressant ficacy of the SSRIs fluoxetine and par- size, calculated by WMD or SMD, and group. On more detailed review, an ad- oxetine in reducing pain (SMD, −0.39; designated a D greater than 0.2 ditional 16 papers were excluded for the 95% CI, −0.77 to −0.01; P=.04). The ef- through 0.5 as a small effect size, a D following reasons: no recognized cri- fects were small on depressed mood greater than 0.5 up to 0.8 as a medium teria for FMS,30 duplicate publica- (WMD, −0.37; 95% CI, −0.66 to −0.07, effect size, and a D greater than 0.8 as tion,31 report of N-of-1 trials,32 lack of P=.02) and HRQOL (WMD, −0.41; a large effect size. We used the follow- an antidepressant-only group,33-35 lack 95% CI, −0.78 to −0.05, P=.03). There ing descriptors to classify meta- of a pharmacological placebo group,36-39 were no effects on fatigue (WMD, analysis results21: “strong” indicated outcome measures not suitable for −0.17; 95% CI, −0.47 to 0.12, P=.25) consistent findings in multiple (at meta-analysis,40 or means or standard or sleep (SMD, −0.23; 95% CI, −0.56 least 2) high- or moderate-quality deviations of pretest and posttest data to 0.10, P=.18). RCTs; “moderate” indicated consistent or standard deviations of change scores There was strong evidence for the ef- findings in multiple low-quality RCTs were not included in the publication ficacy of the SNRIs duloxetine and mil- or 1 high- or moderate-quality RCT; and were not provided by the authors nacipran in reducing pain (SMD, −0.36; “limited” indicated 1 low-quality RCT; on request.41-46 The remaining 18 stud- 95% CI, −0.46 to −0.25; PϽ.001) and and “conflicting” indicated inconsis- ies met our selection criteria and were sleep disturbances (SMD, −0.31; 95% tent findings among multiple RCTs. included in the meta-analysis CI, −0.47 to −0.14; PϽ.001). There was A sensitivity analysis was con- (FIGURE 1).47-64 The interrater reliabil- strong evidence for the efficacy of du- ducted to determine whether different ity for this assessment was ␬=0.92. loxetine in improving depressed mood classes of antidepressants (TCA, SSRI, (SMD, −0.26; 95% CI, −0.42 to −0.10; SNRI, MAOI) influenced the results by Meta-analyses P=.001) and HRQOL (SMD, −0.31; calculating the effect sizes of the out- The effect sizes for all antidepressants 95% CI, −0.44 to −0.17; PϽ.001). Based comes assessed of each class of are shown in FIGURES 2, 3, 4, 5, and 6. on Cohen categories for the magni- antidepressants. There was strong evidence for a reduc- tude of effect size, these effect sizes were

Downloaded From: https://jamanetwork.com/ on 09/28/2021 ANTIDEPRESSANT THERAPY IN FIBROMYALGIA SYNDROME

small. There was no effect of dulox- degree of adverse effects (slight, mod- RCTs studied SNRIs (duloxetine in 3, etine on fatigue (WMD, −0.08; 95% CI, erate, severe).50,59,64 The median fre- milnacipran in 1). Five studies had mul- −0.20 to 0.05; P=.23). quency of severe adverse effects was not tiple groups: Arnold et al50 compared du- There was strong evidence for the ef- different between treatment and pla- loxetine (60 and 120 mg/d) with pla- ficacy of the MAOIs moclobemide and cebo (2.30% vs 2.95%; P=.60). cebo. Russell et al64 compared duloxetine pirlindole in reducing pain (SMD, (20-60 mg/d, 60 mg/d, and 120 mg/d) −0.54; 95% CI, −1.02 to −0.07; P=.03). Validity Analysis with placebo. Goldenberg et al55 com- There was no evidence of efficacy for Characteristics of included studies are pared fluoxetine, amitriptyline, and moclobemide on fatigue (WMD, 0.30; presented in TABLE 2. Interrater reli- the combination of both drugs with 95% CI, −1.04 to 1.64; P=.66) or sleep ability for characteristics shown in placebo. Hannonen et al56 compared disturbances (WMD, 1.00; 95% CI, Table 2 was ␬=0.89. moclobemide and amitriptyline with −0.49 to 2.49; P=.19). There was no Seven studies had a multicenter de- placebo. Heymann et al57 compared effect of pirlindole on depressed mood sign. Eleven had a single-center design. amitriptyline and nortriptyline with (WMD, 0.18; 95% CI, −2.16 to 2.52; Sixteen studies used a parallel design and placebo. P=.88). 2 used a crossover design. Tricyclic or The median duration of the RCTs Median rates of reported adverse ef- tetracyclic antidepressants were inves- was 8 weeks (range, 4-28 weeks). Out- fects (antidepressants, 75.5%, vs pla- tigated in 7 studies (amitriptyline in 7, comes were assessed at the end of the cebo, 62.5%; P=.49) and dropout due nortriptyline in 1), and MAOIs in 3 (mo- treatment. No study measured out- to adverse effects (antidepressants, clobemide in 2, pirlindole in 1). Selec- comes at an additional follow-up visit 15.7%; placebo, 8.1%; P=.18) did not tive serotonin reuptake inhibitors were after treatment cessation. differ between treatment and placebo investigated in 6 studies (fluoxetine in Serum antidepressant levels were groups. Only 3 studies differentiated the 3, citalopram in 2, paroxetine in 1). Four not measured in any RCTs to assess

Figure 2. Effectiveness of Antidepressants in Fibromyalgia for the Outcome Pain

Treatment Control

Individuals, Effect Size, Individuals, Effect Size, SMD Favors Favors Study No. Mean (SD) No. Mean (SD) (95% CI) Treatment Control Weight, % Mean Carette et al,51 1986 27 4.30 (3.00) 32 5.00 (3.00) –0.23 (–0.74 to 0.28) 3.96 Carette et al,52 1995 22 5.07 (3.22) 22 7.13 (2.41) –0.71 (–1.32 to –0.10) 3.10 Ginsberg et al,53 1998 33 4.85 (2.11) 28 6.79 (1.53) –1.03 (–1.56 to –0.49) 3.72 Ginsberg et al,54 1996 20 3.80 (2.40) 20 7.00 (1.30) –1.63 (–2.35 to –0.90) 2.38 Goldenberg et al,55 1996 22 5.75 (2.57) 19 8.15 (1.65) –1.07 (–1.73 to –0.41) 2.75 Goldenberg et al,55 1996 21 6.40 (2.83) 19 8.15 (1.65) –0.73 (–1.37 to –0.09) 2.87 Hannonen et al,56 1998 30 4.50 (2.70) 30 5.20 (2.70) –0.26 (–0.76 to 0.25) 4.02 Hannonen et al,56 1998 32 4.50 (2.80) 30 5.20 (2.70) –0.25 (–0.75 to 0.25) 4.10 Kempenaers et al,63 1994 6 3.20 (3.10) 8 3.70 (2.80) –0.16 (–1.22 to 0.90) 1.24 Wolfe et al,61 1994 15 1.60 (0.79) 9 1.60 (0.79) 0.00 (–0.83 to 0.83) 1.92 Yavuzer et al,62 1998 26 1.57 (0.88) 22 1.88 (0.83) –0.36 (–0.93 to 0.22) 3.41 Subtotal 254 239 –0.59 (–0.86 to –0.31) 33.47 χ 2 2 Test for heterogeneity: 10 = 21.23; P = .02; I = 52.9% Test for overall effect: z = 4.21; P<.001

Mean change Anderberg et al,47 2000 17 –1.00 (1.86) 18 0.00 (2.47) –0.45 (–1.12 to 0.23) 2.68 Arnold et al,50 2005 116 –2.39 (2.37) 118 –1.16 (2.28) –0.53 (–0.79 to –0.27) 7.90 Arnold et al,50 2005 114 –2.40 (2.35) 118 –1.16 (2.28) –0.53 (–0.80 to –0.27) 7.87 Arnold et al,49 2004 101 –1.98 (3.01) 103 –1.35 (2.94) –0.21 (–0.49 to 0.06) 7.60 Arnold et al,48 2002 19 –2.30 (2.40) 18 –0.10 (2.50) –0.88 (–1.56 to –0.20) 2.64 Norregaard et al,58 1995 21 1.00 (2.10) 21 0.70 (1.10) 0.18 (–0.43 to 0.78) 3.14 Patkar et al,59 2007 38 –12.20 (18.50) 48 –8.80 (16.60) –0.19 (–0.62 to 0.23) 5.00 Russell et al,64 2008 79 –2.22 (2.49) 144 –1.43 (2.52) –0.31 (–0.59 to –0.04) 7.58 Russell et al,64 2008 150 –1.98 (2.57) 144 –1.43 (2.52) –0.22 (–0.44 to 0.01) 8.57 Russell et al,64 2008 147 –2.26 (2.55) 144 –1.43 (2.52) –0.33 (–0.56 to –0.10) 8.52 Vitton et al,60 2004 97 –2.30 (3.00) 28 –0.90 (2.90) –0.47 (–0.89 to –0.04) 5.03 Subtotal 899 904 –0.35 (–0.46 to –0.24) 66.53 χ 2 2 Test for heterogeneity: 10 = 12.19; P = .27; I = 17.9% Test for overall effect: z = 6.31; P<.001

Overall 1153 1143 –0.43 (–0.55 to –0.30) 100.00 χ 2 2 Test for heterogeneity: 21 = 37.73; P = .01; I = 44.3% Test for overall effect: z = 6.69; P<.001 –3 –2 –1 0 1 SMD (95% CI)

CI indicates confidence interval; SMD, standardized mean difference.

Downloaded From: https://jamanetwork.com/ on 09/28/2021 ANTIDEPRESSANT THERAPY IN FIBROMYALGIA SYNDROME

Figure 3. Effectiveness of Antidepressants in Fibromyalgia for the Outcome Fatigue

Treatment Control

Individuals, Effect Size, Individuals, Effect Size, SMD Favors Favors Study No. Mean (SD) No. Mean (SD) (95% CI) Treatment Control Weight, % Mean Carette et al,52 1995 22 5.62 (3.07) 20 7.64 (1.80) –0.78 (–1.41 to –0.15) 3.52 Ginsberg et al,54 1996 20 3.80 (2.50) 20 5.90 (2.20) –0.87 (–1.53 to –0.22) 3.31 Goldenberg et al,55 1996 22 6.86 (2.41) 19 7.37 (2.51) –0.20 (–0.82 to 0.41) 3.68 Goldenberg et al,55 1996 21 6.77 (2.99) 19 7.37 (2.51) –0.21 (–0.83 to 0.41) 3.61 Hannonen et al,56 1998 30 4.90 (2.70) 30 4.60 (2.60) 0.11 (–0.39 to 0.62) 5.23 Hannonen et al,56 1998 32 4.70 (2.80) 30 4.60 (2.60) 0.04 (–0.46 to 0.53) 5.39 Wolfe et al,61 1994 15 7.70 (3.98) 9 7.80 (4.20) –0.02 (–0.85 to 0.80) 2.12 Subtotal 162 147 –0.26 (–0.55 to 0.04) 26.87 χ 2 2 Test for heterogeneity: 6 = 9.75; P = .14; I = 38.5% Test for overall effect: z = 1.71; P = .09

Mean change Anderberg et al,47 2000 17 –0.63 (3.01) 18 –0.22 (2.42) –0.15 (–0.81 to 0.52) 3.20 Arnold et al,49 2004 101 –1.30 (2.91) 103 –0.88 (2.84) –0.15 (–0.42 to 0.13) 13.89 Arnold et al,48 2002 19 –1.60 (2.80) 18 0.40 (2.80) –0.70 (–1.37 to –0.03) 3.18 Norregaard et al,58 1995 21 0.50 (2.20) 21 0.10 (2.00) 0.19 (–0.42 to 0.79) 3.78 Russell et al,64 2008 79 –1.79 (3.91) 144 –1.69 (4.08) –0.02 (–0.30 to 0.25) 13.92 Russell et al,64 2008 150 –1.83 (4.16) 144 –1.69 (4.08) –0.03 (–0.26 to 0.19) 17.64 Russell et al,64 2008 147 –2.12 (4.00) 144 –1.69 (4.08) –0.11 (–0.34 to 0.12) 17.52 Subtotal 534 592 –0.09 (–0.21 to 0.03) 73.13 χ 2 2 Test for heterogeneity: 6 = 4.66; P = .59; I = 0% Test for overall effect: z = 1.46; P = .14

Overall 696 739 –0.13 (–0.26 to –0.01) 100.00 χ 2 2 Test for heterogeneity: 13 = 15.66; P = .27; I = 17.0% Test for overall effect: z = 2.10; P = .04 –2 –1 0 1 SMD (95% CI)

CI indicates confidence interval; SMD, standardized mean difference.

Figure 4. Effectiveness of Antidepressants in Fibromyalgia for the Outcome Sleep

Treatment Control

Individuals, Effect Size, Individuals, Effect Size, SMD Favors Favors Study No. Mean (SD) No. Mean (SD) (95% CI) Treatment Control Weight, % Mean Carette et al,52 1995 22 3.93 (3.14) 20 6.51 (2.69) –0.86 (–1.50 to –0.23) 4.61 Ginsberg et al,54 1996 24 2.60 (3.10) 22 5.10 (3.00) –0.80 (–1.41 to –0.20) 5.08 Goldenberg et al,55 1996 22 6.66 (2.66) 19 7.46 (2.39) –0.31 (–0.93 to 0.31) 4.86 Goldenberg et al,55 1996 21 5.70 (3.48) 19 7.46 (2.39) –0.57 (–1.21 to 0.06) 4.62 Hannonen et al,56 1998 30 5.80 (3.00) 30 4.80 (2.90) 0.33 (–0.18 to 0.84) 6.93 Hannonen et al,56 1998 32 3.60 (2.80) 39 4.80 (2.90) –0.42 (–0.89 to 0.06) 7.94 Kempenaers et al,63 1994 6 4.70 (3.00) 8 6.00 (2.20) –0.47 (–1.55 to 0.60) 1.67 Wolfe et al,61 1994 15 7.60 (3.10) 9 7.60 (3.83) 0.00 (–0.83 to 0.83) 2.80 Subtotal 172 166 –0.38 (–0.68 to –0.07) 38.49 χ 2 2 Test for heterogeneity: 7 = 12.90; P = .07; I = 45.7% Test for overall effect: z = 2.42; P = .02

Mean change Anderberg et al,47 2000 17 –0.59 (0.62) 18 –0.39 (1.80) –0.14 (–0.81 to 0.52) 4.25 Arnold et al,50 2005 116 –2.67 (3.12) 118 –1.71 (3.04) –0.31 (–0.57 to –0.05) 21.35 Arnold et al,50 2005 114 –2.69 (3.09) 118 –1.71 (3.04) –0.32 (–0.58 to –0.06) 21.21 Norregaard et al,58 1995 21 –1.00 (2.90) 21 –0.10 (2.50) –0.33 (–0.94 to 0.28) 4.99 Vitton et al,60 2004 97 –1.30 (3.10) 28 –0.50 (2.90) –0.26 (–0.68 to 0.16) 9.71 Subtotal 365 303 –0.30 (–0.46 to –0.14) 61.51

χ 2 2 Test for heterogeneity: 4 = 0.28; P = .99; I = 0% Test for overall effect: z = 3.72; P<.001

Overall 537 469 –0.32 (–0.46 to –0.18) 100.00 χ 2 2 Test for heterogeneity: 12 = 13.38; P = .34; I = 10.3% Test for overall effect: z = 4.47; P<.001 –2 –1 0 1 SMD (95% CI)

CI indicates confidence interval; SMD, standardized mean difference.

Downloaded From: https://jamanetwork.com/ on 09/28/2021 ANTIDEPRESSANT THERAPY IN FIBROMYALGIA SYNDROME

patients’ adherence. All RCTs acetaminophen in combination with age of the allowed comedication. No allowed additional therapy with acetylsalicylic acid or nonsteroidal study controlled for additional paracetamol or acetaminophen. Eight anti-inflammatory drugs or codeine. therapy with comedication. No study allowed therapy with paracetamol or Seven studies reported a defined dos- provided detailed information about

Figure 5. Effectiveness of Antidepressants in Fibromyalgia for the Outcome Depressed Mood

Treatment Control

Individuals, Effect Size, Individuals, Effect Size, SMD Favors Favors Study No. Mean (SD) No. Mean (SD) (95% CI) Treatment Control Weight, % Mean Goldenberg et al,55 1996 22 7.80 (4.70) 19 9.30 (6.50) –0.26 (–0.88 to 0.35) 4.62 Goldenberg et al,55 1996 20 8.70 (6.00) 19 9.30 (6.50) –0.09 (–0.72 to 0.53) 4.45 Wolfe et al,61 1994 15 8.30 (5.86) 9 13.90 (10.82) –0.67 (–1.53 to 0.18) 2.42 Yavuzer et al,62 1998 28 5.54 (5.56) 25 5.36 (2.68) 0.04 (–0.50 to 0.58) 6.04 Subtotal 85 72 –0.17 (–0.49 to 0.14) 17.54 χ 2 2 Test for heterogeneity: 3 = 2.07; P = .56; I = 0% Test for overall effect: z = 1.07; P = .29

Mean change Anderberg et al,47 2000 17 –1.13 (2.87) 18 –0.33 (2.14) –0.31 (–0.98 to 0.36) 3.95 Arnold et al,50 2005 111 –3.79 (4.34) 109 –2.24 (4.70) –0.34 (–0.61 to –0.08) 24.80 Arnold et al,50 2005 110 –2.97 (4.72) 109 –2.24 (4.70) –0.15 (–0.42 to 0.11) 24.98 Arnold et al,49 2004 88 –3.32 (7.97) 89 –1.02 (7.83) –0.29 (–0.59 to 0.01) 20.04 Arnold et al,48 2002 19 –1.50 (2.80) 18 0.50 (2.10) –0.79 (–1.46 to –0.12) 3.89 Norregaard et al,58 1995 21 –1.00 (6.10) 21 –0.90 (7.90) –0.01 (–0.62 to 0.59) 4.81 Subtotal 366 364 –0.27 (–0.42 to –0.13) 82.46

χ 2 2 Test for heterogeneity: 5 = 4.00; P = .55; I = 0% Test for overall effect: z = 3.66; P<.001

Overall 451 436 –0.26 (–0.39 to –0.12) 100.00 χ 2 2 Test for heterogeneity: 9 = 6.39; P = .70; I = 0% Test for overall effect: z = 3.77; P<.001 –2 –1 0 1 SMD (95% CI)

CI indicates confidence interval; SMD, standardized mean difference.

Figure 6. Effectiveness of Antidepressants in Fibromyalgia for the Outcome Health-Related Quality of Life

Treatment Control

Individuals, Effect Size, Individuals, Effect Size, SMD Favors Favors Study No. Mean (SD) No. Mean (SD) (95% CI) Treatment Control Weight, % Mean Goldenberg et al,55 1996 2247.60 (19.80) 19 58.50 (17.10) –0.57 (–1.20 to 0.05) 2.88 Goldenberg et al,55 1996 2152.30 (22.90) 19 58.50 (17.10) –0.30 (–0.92 to 0.33) 2.91 Heymann et al,57 2001 3739.97 (19.88) 33 51.68 (22.90) –0.54 (–1.02 to –0.06) 4.74 Heymann et al,57 2001 36 49.78 (21.84) 33 51.68 (22.90) –0.08 (–0.56 to 0.39) 4.85 Subtotal 116 104 –0.36 (–0.62 to –0.09) 15.38 χ 2 2 Test for heterogeneity: 3 = 2.35; P = .50; I = 0% Test for overall effect: z = 2.61; P = .009

Mean change Arnold et al,50 2005 114 –16.72 (16.34) 115 –8.35 (16.40) –0.51 (–0.77 to –0.25) 12.58 Arnold et al,50 2005 112 –16.81 (16.30) 115 –8.35 (16.40) –0.52 (–0.78 to –0.25) 12.50 Arnold et al,49 2004 101 –13.46 (18.29) 102 –7.93 (17.47) –0.31 (–0.58 to –0.03) 11.70 Arnold et al,48 2002 19 –11.50 (14.80) 18 –0.40 (15.00) –0.73 (–1.40 to –0.06) 2.56 Norregaard et al,58 1995 21 0.00 (0.40) 21 0.00 (0.40) 0.00 (–0.60 to 0.60) 3.09 Russell et al,64 200879 –14.77 (16.71) 144 –10.42 (17.52) –0.25 (–0.53 to 0.02) 11.78 Russell et al,64 2008 150 –12.28 (17.64) 144 –10.42 (17.52) –0.11 (–0.33 to 0.12) 15.27 Russell et al,64 2008147 –13.86 (17.10) 144 –10.42 (17.52) –0.20 (–0.43 to 0.03) 15.13 Subtotal 743 803 –0.31 (–0.44 to –0.18) 84.62

χ 2 2 Test for heterogeneity: 7 = 11.16; P = .13; I = 37.3% Test for overall effect: z = 4.54; P<.001

Overall 859 907 –0.31 (–0.42 to –0.20) 100.00 χ 2 2 Test for heterogeneity: 11 = 13.66; P = .25; I = 19.5% Test for overall effect: z = 5.59; P<.001 –2 –1 0 1 SMD (95% CI)

CI indicates confidence interval; SMD, standardized mean difference.

Downloaded From: https://jamanetwork.com/ on 09/28/2021 ANTIDEPRESSANT THERAPY IN FIBROMYALGIA SYNDROME

nonpharmacological therapies or in western Europe (Scandinavia, Bel- for participants randomized to placebo controlled for nonpharmacological gium), and 1 was performed in Tur- (P=.78). The median age of study par- therapies. key. All studies were outpatient- ticipants was 47.0 years. Seven studies Ten studies performed a power based. Patients were recruited from included only women; 10 included both analysis to ensure an adequate sample rheumatology departments in 11 stud- men and women, and 1 study did not size. Five studies had a Jadad score of ies, from research centers in 3 studies, specify participant sex. The median per- 5, 8 had a Jadad score of 4, and 5 had a and from a psychiatric department in centage of women in all studies was 98%. Jadad score less than 4. Eleven studies 1 study. Three publications did not re- Ten studies reported the participants’ had a van Tulder score of 5 to 7, and 7 port the recruitment setting. race. Among these, the median percent- studies had scores of 8 to 11. Only 4 All studies excluded patients with age of white participants was 89.5%. studies were high on both mea- severe somatic diseases. Eleven excluded Fibromyalgia syndrome was defined in sures.49,56,57,59 Interrater reliability for patients with severe mental disorders. 16 studies according to the ACR crite- this assessment was ␬=0.89. Ten excluded specific age categories ria.2 One study defined FMS according There was significant heterogeneity (Ͻ18 or Ͼ60 years). Four excluded to Yunus,19 and 1 study defined FMS between the analyzed RCTs in most patients with pending applications for according to Smythe and Moldofsky.18 outcome measures. The large ranges of disability. A total of 1427 individuals No study provided data on nonpsychi- the 95% CIs are also indicative of completed treatment. Of these, 916 were atric comorbidities. Five studies pro- marked variations between the studies. receiving antidepressants. The median vided data on the prevalence of major Eleven studies were performed in percentages of patients completing the depressive disorder (up to 35%). Three North America or Puerto Rico, 1 was trials were 71.0% for participants ran- studies provided information on work- performed in Brazil, 5 were performed domized to antidepressants and 78.0% ing status (working, sick leave, or disability). Potential participants consid- Table 1. Effect Sizes of the Different Classes of Antidepressants on Selected Outcome ered treatment refractory were excluded Variables in 2 studies. Test for Patients Taking Overall Publication Bias No. of Antidepressants, Statistical Effect Outcome Studies No. Method Effect Size (95% CI) P Value Visual scanning of forest plots for sub- Tricyclic Antidepressants group analysis suggested a random dis- Pain 6 128 SMD (random) −1.64 (−2.57 to −0.71) Ͻ.001 tribution with results in the same di- Fatigue 4 95 SMD (random) −1.12 (−1.87 to −0.38) .003 rection for most outcomes, indicating Sleep 5 105 WMD (fixed) −1.84 (−2.62 to −1.06) Ͻ.001 that although study effect sizes dif- Depressed 1 20 WMD (fixed) −0.60 (−4.53 to 3.33) .76 fered, results were mostly consistent mood (data not shown). The fail-safe num- HRQOL 3 94 WMD (fixed) −0.31 (−0.60 to −0.01) .04 ber with a D=0.01 as the selected cri- Selective Serotonin Reuptake Inhibitors Pain 6 132 SMD (random) −0.39 (−0.77 to −0.01) .04 terion value to “nullify” the average Fatigue 5 94 WMD (fixed) −0.17 (−0.47 to 0.12) .25 effect on pain was n=924; on fatigue, Sleep 4 75 SMD (random) −0.23 (−0.56 to 0.10) .18 n=168; on sleep, n=403; on de- Depressed 5 94 WMD (fixed) −0.37 (−0.66 to −0.07) .02 pressed mood, n=250; and on HRQOL, mood n=360. Thus the fail-safe numbers were HRQOL 3 62 WMD (fixed) −0.41 (−0.78 to −0.05) .03 larger than Rosenthal rule of thumb28 Serotonin and Noradrenaline Reuptake Inhibitors of n=120 for pain, n=80 for fatigue, Pain 3 804 SMD (random) −0.36 (−0.46 to −0.25) Ͻ.001 n=75 for sleep, n=60 for depressed Fatigue 1 477 WMD (fixed) −0.08 (−0.20 to 0.05) .23 mood, and n=70 for HRQOL. These re- Sleep 2 327 SMD (random) −0.31 (−0.47 to −0.14) Ͻ.001 sults indicate that a publication bias is Depressed 2 309 SMD (random) −0.26 (−0.42 to −0.10) .001 unlikely to change the overall results mood of this meta-analysis. HRQOL 2 703 SMD (random) −0.31 (−0.44 to −0.17) Ͻ.001 Monoamine Oxidase Inhibitors COMMENT Pain 3 89 SMD (random) −0.54 (−1.02 to −0.07) .03 Fatigue 1 30 WMD (fixed) 0.30 (−1.04 to 1.64) .66 The primary aim of this meta-analysis Sleep 1 30 WMD (fixed) 1.00 (−0.49 to 2.49) .19 was to determine the efficacy of anti- Depressed 1 28 WMD (fixed) 0.18 (−2.16 to 2.52) .88 depressants for treatment of FMS. We mood found strong evidence for the efficacy HRQOL NA NA NA NA NA of antidepressants in reducing pain, Abbreviations: CI, confidence interval; HRQOL, health-related quality of life; NA, not assessed; SMD, standardized mean sleep disturbances, and depressed mood difference; WMD, weighted mean difference. and for improving HRQOL. All effect

Downloaded From: https://jamanetwork.com/ on 09/28/2021 ANTIDEPRESSANT THERAPY IN FIBROMYALGIA SYNDROME

sizes were small. We found strong evi- We found large effect sizes of TCAs reducing pain; small effect sizes of dence against a favorable effect of an- for reducing pain, fatigue, and sleep dis- SNRIs for reducing pain, sleep distur- tidepressants on fatigue. turbances; small effect sizes of SSRIs for bances, and depressed mood; and small

Table 2. Main Study Characteristics Study Population Treatment Group Placebo Method Group, Quality, Female Com- Com- Com- Jadad Outcome Sex/ Age, Randomized/ pleted, Duration, pleted, pleted, Score Measures Source White Mean, Exclusion Screened, No./ Treatment, No./ No./ (van Tulder Used for (Location) Race, % y Criteria No. (%) Total (%) and Design Total (%) Total (%) Score) Meta-analysis Tricyclic Antidepressants: Amitriptyline Carette et al,51 92.6/NR 41.8 SSD, IRD NR 59/70 9 wk, amitriptyline 27/34 32/36 4 (9) Pain, VAS 1986 (84.3) 50 mg/d, (79.4) (88.9) (Canada) parallel Kempenaers et 100/NR 38.7 SSD, IRD, mental NR 14/24 8 wk, amitriptyline 6/12 8/12 3 (7) Pain, VAS; al,63 1994 disorder (58.3) 50 mg/d, (50.0) (75.0) sleep, VAS (Belgium) parallel Carette et al,52 95.5/NR 43.8 SSD, IRD NR 20/22 8 wk, amitriptyline 20/22 20/22 4 (8) Pain, VAS; 1995 (90.9) 25 mg/d, (90.9) (90.9) sleep, VAS (Canada) crossover Ginsberg et 83/92 46 Age, SSD, IRD NR 46/51 8 wk, 24/26 22/25 4 (8) Pain, VAS; fatigue, al,54 1996 (90.2) sustained- (92.3) (88) VAS; sleep, (Belgium) release VAS amitriptyline 25 mg/d, parallel Goldenberg et 90/100 43.2 Age, somatic NR 19/31 6 wk, amitriptyline 19/31 NR 5 (7) Pain, VAS; al,55 1996 disease, major (61.3) 25 mg/d, (61.3) fatigue, VAS; (United depression crossover sleep, VAS; States) depression, BDI; quality of life, FIQ total score Hannonen et 100/NR 49.7 Age, SSD, severe 130/184 92/130 12 wk, amitriptyline 32/42 30/45 5 (10) Pain, VAS; fatigue, al,56 1998 mental disorder (70.6) (70.8) 12.5 mg/d, (76.2) (66.7) VAS; sleep, (Finland) parallel VAS; quality of life, NHP Heymann et 100/65 53.4 Age, SSD, IRD NR 106/118 8 wk, amitriptyline 37/40 33/40 5 (8) Quality of life, FIQ al,57 2001 (89.8) 25 mg/d, (92.5) (82.5) total score (Brazil) parallel Tricyclic Antidepressants: Nortriptyline Heymann et 100/65 53.4 Age, SSD, IRD NR 106/118 8 wk, nortriptyline 36/38 33/40 5 (8) Quality of life, FIQ al,57 2001 (89.8) 25 mg/d, (94.7) (82.5) total score (Brazil) parallel Selective Serotonin Reuptake Inhibitors: Paroxetine Patkar et al,59 94/NR 47.9 Age, SSD, IRD, 116/983 86/116 12 wk, paroxetine 38/58 48/58 5 (9) Pain, VAS 2007 mental disorder, (11.8) (74.1) controlled (65.5) (82.8) (United pending release 62.5 States) disability review mg/d (mean, 39.1 mg/d), parallel Selective Serotonin Reuptake Inhibitors: Fluoxetine Wolfe et al,61 100/100 48 IRD NR 24/42 6 wk, fluoxetine 20 15/21 9/21 3 (7) Pain, VAS; 1994 (57.1) mg/d, parallel (71.4) (42.8) fatigue, VAS; (United sleep, VAS; States) depression, BDI Goldenberg et 90/100 43.2 Age, somatic NR 19/31 6 wk, fluoxetine 20 NA 19/31 5 (7) Pain, VAS; al,55 1996 disease, major (61.3) mg/d, (61.3) fatigue, VAS; (United depression crossover sleep, VAS; States) depression, BDI; quality of life, FIQ total score Arnold et al,48 100/90 46 Age, SSD, IRD, NR 37/60 12 wk, fluoxetine 19/30 18/30 4 (7) Pain, FIQ; 2002 mental disorder (61.7) 20-80 mg/d, (63.3) (60) fatigue, FIQ; (United parallel depression, States) FIQ; quality of life, FIQ total score (continued)

Downloaded From: https://jamanetwork.com/ on 09/28/2021 ANTIDEPRESSANT THERAPY IN FIBROMYALGIA SYNDROME

Table 2. Main Study Characteristics (continued) Study Population Treatment Group Placebo Method Group, Quality, Female Com- Com- Com- Jadad Outcome Sex/ Age, Randomized/ pleted, Duration, pleted, pleted, Score Measures Source White Mean, Exclusion Screened, No./ Treatment, No./ No./ (van Tulder Used for (Location) Race, % y Criteria No. (%) Total (%) and Design Total (%) Total (%) Score) Meta-analysis Selective Serotonin Reuptake Inhibitors: Citalopram Nørregaard et NR/NR 48 SSD, IRD, mental 42/150 (28) 33/42 8 wk, citalopram 12/21 21/21 4 (6) Pain, NRS; al,58 1995 disorder (78.6) 20-40 mg/d, (57.1) (100) fatigue, NRS; (Denmark) parallel sleep, NRS; depression, BDI; quality of life, FIQ physical score Anderberg et 100/NR 48.6 SSD, major NR 35/40 16 wk, citalopram 17/21 18/19 4 (9) Pain, FIQ; fatigue, al,47 2000 depression (87.5) 20-40 mg/d, (80.9) (95) FIQ; sleep, (Sweden) parallel MADRS; depression, FIQ Serotonin and Noradrenaline Reuptake Inhibitors: Duloxetine Arnold et al,49 88.5/ 49.9 Age, SSDs, IRD, 207/555 124/207 12 wk, duloxetine 58/104 66/103 5 (8) Pain, FIQ; 2004 88.5 mental disorder (37.3) (59.9) 120 mg/d, (55.7) (64.1) fatigue, FIQ; (United except major parallel depression, States) depressive BDI; quality of disorder, life, FIQ total pending score disability review Arnold et al,50 100/89.5 49.6 Age, SSD, IRD, 354/745 215/354 12 wk, duloxetine 147/234 68/120 3 (7) Pain, BPI; 2005 mental disorder (47.5) (60.7) 60 mg/d or 120 (61.5) (56.7) sleep, BPI; (United except major mg/d, parallel depression, States) depressive HDRS; quality disorder, of life, FIQ total pending score disability reviews, patients judged refractory to treatment Russell et al,64 94.8/ 51.0 Age, SSD, IRD, 520/1010 323/520 28 wk, duloxetine 251/376 72/144 4 (8) Pain, BPI; fatigue, 2008 84.2 mental disorder (51.8) (62.1) 20 mg/d or 60 (66.7) (50) MFI; quality of (United except major mg/d or 120 life, FIQ total States) depressive mg/d, parallel score disorder, pending disability reviews, patients judged refractory to treatment Serotonin and Noradrenaline Reuptake Inhibitors: Milnacipran Vitton et al,60 96-98/ 46-48 Age, somatic 125/184 90/125 12 wk, milnacipran 69/97 21/28 3 (7) Pain, VAS; sleep, 2004, 79-89 disease, (67.9) (72) 100 mg/d or (71.1) (75) Jenkins scale (United severe 200 mg/d, States) mental parallel disorder Monoamine Oxidase Inhibitors: Moclobemide Hannonen et 100/NR 49.7 Age, SSD, severe 130/184 92/130 12 wk, 30/43 30/45 5 (10) Pain, VAS; al,56 1998 mental disorder (70.6) (70.8) moclobemide (69.8) (66.7) fatigue, VAS; (Finland) 150 mg/d, sleep, VAS parallel Yavuzer et al,62 58.3/NR 33.2 SSD, IRD NR 53/60 6 wk, moclobemide 26/28 22/25 1 (6) Pain, NRS; 1998 (88.3) 300 mg/d, (92.9) (88) depression, (Turkey) parallel HDRS Monoamine Oxidase Inhibitors: Pirlindole Ginsberg et 87.9/NR 39.7 Age, somatic 100/200 61/100 4 wk, pirlindole 150 33/50 28/50 3 (7) Pain, VAS; al,53 1998 disease (50) (61) mg/d, parallel (66) (56) fatigue, VAS; (Belgium) sleep, VAS Abbreviations: BDI, Beck Depression Inventory; BPI, Brief Pain Inventory; FIQ, Fibromyalgia Impact Questionnaire; HDRS, Hamilton Depression Rating Scale; IRD, inflammatory rheu- matologic disease; MADRS, Montgomery Asberg Depression Rating Scale; MFI, Multidimensional Fatigue Inventory; NHP, Nottingham Health Profile; NA, not assessed; NR, not reported; NRS, Numeric Rating Scale; SSD, severe somatic disease; VAS, visual analog scale.

Downloaded From: https://jamanetwork.com/ on 09/28/2021 ANTIDEPRESSANT THERAPY IN FIBROMYALGIA SYNDROME

effect sizes of MAOIs for reducing pain. The external validity of the RCTs studies with TCAs, 3 studies with Conclusions regarding the efficacy of analyzed was limited by the following. SSRIs, and 2 studies with S-adenosyl- single drugs on outcomes were lim- First, the short duration of most stud- methionine. Medium effect sizes were ited because of small sample sizes. Of ies and the lack of follow-up after reported for improving pain, sleep, and the antidepressants studied, dulox- treatment cessation leave unanswered fatigue. Sensitivity analysis by way of etine is the only one that has been ap- whether antidepressants have long- meta-regression revealed no effect of proved by the Food and Drug Admin- term beneficial effects on FMS symp- drug class.13 Differences between re- istration for treating FMS.65 toms and the optimal treatment dura- sults of the meta-analysis of O’Malley Three studies that did not meet in- tion. One excluded RCT failed to et al and our review regarding class ef- clusion criteria for this meta-analysis demonstrate an advantage of amitrip- fects are due to the fact that we ana- were head-to-head comparisons of an- tyline over placebo regarding pain, lyzed more studies with TCAs and SS- tidepressant drugs. Of these, 1 study sleep disturbances, and fatigue after RIs and included RCTs of MAOIs and found that paroxetine was superior to 26 weeks.43 Second, despite evidence SNRIs. In addition, the studies failing amitriptyline in reducing pain,41 while of higher prevalences of mental disor- to demonstrate an advantage of citalo- another study found that amitripty- ders in FMS,10 only 6 studies per- pram over placebo47,58 were published line was superior to paroxetine in re- formed a standardized psychiatric after the meta-analysis by O’Malley et ducing pain and sleep disturbances.42 interview. Only 3 studies performed al. Arnold et al14 included 9 studies (6 A third study found no difference be- subgroup analyses among participants with TCAs and 3 with cyclobenzap- tween amitriptyline and fluoxetine in with vs without major depressive rine) into a meta-analysis and found a reducing sleepiness.40 Overall, none of disorder. The effects of duloxetine global effect size of 0.44 for pain, sleep, these head-to-head comparisons of dif- on pain did not differ between tenderness, fatigue, and sleep. ferent antidepressant classes nor this FMS patients with vs without major This review has limitations. First, meta-analysis allows a definitive con- depressive disorder.49,50,64 Therefore, since demographics and comorbidi- clusion regarding superiority of one only duloxetine has demonstrated ties of study participants and the class of antidepressants over another. efficacy in FMS patients both with amount of comedication were not Doses of TCAs used in the studies, and without major depressive disor- reported, these possible sources of between 12.5 and 50 mg per day, were der. Third, no definitive statements heterogeneity could not be examined. typical for pain treatment but far be- are possible on the efficacy of antide- Second, we did not seek to identify low the doses of TCAs necessary for an pressants in men, nonwhite individu- unpublished studies. Third, 6 RCTs antidepressant benefit. This likely ex- als, patients older than 65 years, chil- were excluded because the published plains the positive association of TCAs dren, and adolescents because these data were not suitable for meta- for reducing pain in the absence of a subgroups were not analyzed, with analysis, and necessary data were not benefit for depressive symptoms. In the exception of 2 studies with dulox- provided by the authors on request. contrast, doses of SSRIs and SNRIs were etine: 1 study found no significant With the exception of 1 study, which equal to those used for treating affec- response in primary and secondary failed to demonstrate an advantage of tive disorders. However, we could only outcomes in male patients.49 Another amitriptyline over placebo regarding find an effect of SNRIs on depressed study reported similar significant pain pain, sleep disturbances, and fatigue mood. reduction between women and men.64 after 26 weeks,43 the other 5 studies The internal validity of the RCTs ana- One study reported similar results excluded from the meta-analysis lyzed was limited for the following rea- among racial groups and patients aged reported an advantage of amitripty- sons. First, serum antidepressant lev- 65 years and younger.64 Fourth, since line41,42,45,46 or paroxetine41,42,44 over els were not measured in any RCTs to most studies excluded patients with placebo regarding pain and sleep dis- assess patients’ adherence. Second, no severe somatic diseases, including turbances that was consistent with study controlled for consumption, dose, inflammatory arthritic diseases, it is our results. Fourth, there are limita- or adverse effects of concomitant an- unknown whether antidepressants are tions of some methods used in this algesic medications. The influence of effective in these patients with FMS. article, such as using I2 for assessing this comedication on study outcomes Finally, 2 duloxetine studies excluded the amount of heterogeneity in is unclear. Third, 3 studies of dulox- patients who were judged by the random-effects meta-analysis66,67 and etine used a 1-week, single-blind, pla- investigator to be treatment refrac- fail-safe numbers68 for excluding a cebo-lead-in phase. Medication ad- tory.50,64 This procedure could have publication bias. verse effects indicating the presence of favored the outcomes of the treatment active drug may have biased the dulox- group. CONCLUSION etine trials more than other studies. Fi- Our findings are mainly consistent Short-term usage of amitriptyline and nally, some studies did not report the with published literature. The meta- duloxetine can be considered for the results of all outcomes assessed. analysis of O’Malley et al13 included 9 treatment of pain and sleep distur-

Downloaded From: https://jamanetwork.com/ on 09/28/2021 ANTIDEPRESSANT THERAPY IN FIBROMYALGIA SYNDROME

bances in FMS. This recommendation from the German Rheumatology League and Ger- understanding of the ‘fibrositis’ syndrome. Bull Rheum man Fibromyalgia Association, German Ministry of Dis. 1977;28(1):928-931. is based on the number of patients stud- Education and Research (Bundesministerium für Bil- 19. Yunus MB. Primary fibromyalgia syndrome: a criti- ied (duloxetine) and on the effect sizes dung und Forschung), German Research Network on cal evaluation of recent criteria developments. Neuropathic Pain, and intramural funds of the Uni- Z Rheumatol. 1989;48(5):217-222. (amitriptyline). Before treatment is ini- versity of Würzburg (Germany). 20. Mease PJ, Arnold LM, Crofford LJ, et al. Identi- tiated, concomitant diseases related to Role of the Sponsor: The sponsors had no role in the fying the clinical domains of fibromyalgia: contribu- potential adverse effects of the drugs design and conduct of the study; in the collection, tions from clinician and patient Delphi exercises. analysis, and interpretation of the data; or in the prepa- Arthritis Rheum. 2008;59(7):952-960. and patients’ preferences should be con- ration, review, or approval of the manuscript. 21. van Tulder M, Furlan A, Bombardier C, Bouter L; sidered. Goals of pharmacological Editorial Board of the Cochrane Collaboration Back Review Group. Updated method guidelines for therapy should be defined (no cure, but REFERENCES systematic reviews in the Cochrane Collaboration possible symptom reduction). Since evi- Back Review Group. Spine. 2003;28(12):1290- 1. Gran JT. The epidemiology of chronic generalized dence for a long-term effect of antide- 1299. musculoskeletal pain. Best Pract Res Clin Rheumatol. 22. Jadad AR, Moore RA, Carroll D, et al. Assessing pressants in FMS is still lacking, their 2003;17(4):547-561. the quality of reports of randomized clinical trials: is effects should be reevaluated at regu- 2. Wolfe F, Smythe HA, Yunus MB, et al; Report of blinding necessary? Control Clin Trials. 1996;17 the Multicenter Criteria Committee. The American Col- (1):1-12. lar intervals to determine whether ben- lege of Rheumatology 1990 Criteria for the Classifi- 23. Review Manager (RevMan) version 4.2.10 [com- efits outweigh adverse effects. cation of Fibromyalgia. Arthritis Rheum. 1990; puter program]. The Nordic Cochrane Centre, The 33(2):160-172. Cochrane Collaboration; 2003. Studies of longer duration than those 3. Bennett RM, Jones J, Turk DC, Russell IJ, Matallana 24. Altman DG, Bland JM. Standard deviations and currently available are needed to inves- L. An internet survey of 2,596 people with fibromyalgia. standard errors. BMJ. 2005;331(7521):903. BMC Musculoskelet Disord . 2007;8:27. 25. Cohen J. Statistical Power Analysis for the Be- tigate the long-term efficacy of antide- 4. Häuser W, Zimmer C, Felde E, Kollner V. What are havioral Sciences. 2nd ed. Hillsdale, OK: Lawrence the key symptoms of fibromyalgia? results of a sur- pressant therapy for FMS. It is cur- Erlbaum Assoc; 1988. vey of the German Fibromyalgia Association [in 26. Egger M, Davey Smith G, Schneider M, Minder rently unknown whether benefits of German]. Schmerz. 2008;22(2):176-183. C. Bias in meta-analysis detected by a simple, graphi- 5. Wolfe F, Anderson J, Harkness D, et al. Work and antidepressants for treatment of FMS cal test. BMJ. 1997;315(7109):629-634. disability status of persons with fibromyalgia. 27. Orwin RG. A fail-safe N for effect size in persist after cessation of therapy. It is J Rheumatol. 1997;24(6):1171-1178. meta-analysis. J Stat Educat. 1983;8:157-159. 6. Boonen A, van den Heuvel R, van Tubergen A, et al. also unknown whether antidepres- 28. Rosenthal R. The “file drawer problem” and tol- 9 Large differences in cost of illness and wellbeing be- sants reduce FMS-related costs. The erance for null results. Psychol Bull. 1979;86:638- tween patients with fibromyalgia, chronic low back 641. identification of patient characteris- pain, or ankylosing spondylitis. Ann Rheum Dis. 2005; 29. Hayashino Y, Noguchi Y, Fukui T. Systematic evalu- tics associated with positive and nega- 64(3):396-402. 7. Penrod JR, Bernatsky S, Adam V, Baron M, Dayan ation and comparison of statistical tests for publica- tive therapeutic outcomes are needed N, Dobkin PL. Health services costs and their deter- tion bias. J Epidemiol. 2005;15(6):235-243. to better target antidepressant therapy minants in women with fibromyalgia. J Rheumatol. 30. Bibolotti E, Borghi C, Pasculli E. The manage- 2004;31(7):1391-1398. ment of fibrositis: a double-blind comparison of ma- for FMS. Future studies of the effects 8. Henriksson CM, Liedberg GM, Gerdle B. Women protiline (Ludiomil) and placebo. Clin Trials J. 1986; of antidepressants on FMS should in- with fibromyalgia: work and rehabilitation. Disabil 23:269-280. Rehabil. 2005;27(12):685-694. 31. Gendreau RM, Thorn MD, Gendreau JF, et al. Ef- clude patients with somatic and men- 9. Robinson RL, Jones ML. In search of pharmacoeco- ficacy of milnacipran in patients with fibromyalgia. tal comorbidities and fully report all pa- nomic evaluations for fibromyalgia treatments: a J Rheumatol. 2005;32(10):1975-1985. review. Expert Opin Pharmacother. 2006;7(8): 32. Jaeschke R, Adachi J, Guyatt G, Keller J, Wong tient characteristics and outcomes B. Clinical usefulness of amitriptyline in fibromyalgia: 69 1027-1039. assessed. 10. Henningsen P, Zipfel S, Herzog W. Manage- the results of 23 N-of-1 randomized controlled trials. ment of functional somatic syndromes. Lancet. 2007; J Rheumatol. 1991;18(3):447-451. Author Contributions: Dr Häuser had full access to all 369(9565):946-955. 33. Cantini F, Bellandi F, Niccoli L, Di Munno O. Fluox- of the data in the study and takes responsibility for 11. Goldenberg DL, Burckhardt C, Crofford L. Man- etin combined with cyclobenzaprine in the treatment the integrity of the data and the accuracy of the data agement of fibromyalgia syndrome. JAMA. 2004; of fibromyalgia [in Italian]. Minerva Med. 1994; analysis. 292(19):2388-2395. 85(3):97-100. Study concept and design: Häuser, Üçeyler, Sommer. 12. Carville SF, Arendt-Nielsen S, Bliddal H, et al; 34. Isomeri R, Mikkelsson M, Latikka P, Kammonen Acquisition of data: Häuser, Bernardy, Üçeyler, EULAR. EULAR evidence-based recommendations for K. Effects of amitriptyline and cardiovascular fitness Sommer. the management of fibromyalgia syndrome. Ann training on pain in patients with primary fibromyalgia. Analysis and interpretation of data: Häuser, Bernardy, Rheum Dis. 2008;67(4):536-541. J Musculoskel Pain. 1993;1:253-260. Üçeyler, Sommer. 13. O’Malley PG, Balden E, Tomkins G, Santoro J, 35. Fors EA, Sexton H, Gotestam KG. The effect of Drafting of the manuscript: Häuser, Bernardy. Kroenke K, Jackson JL. Treatment of fibromyalgia with guided imagery and amitriptyline on daily fibromyal- Critical revision of the manuscript for important in- antidepressants: a meta-analysis. J Gen Intern Med. gia pain: a prospective, randomized, controlled trial. tellectual content: Häuser, Bernardy, Üçeyler, Sommer. 2000;15(9):659-666. J Psychiatr Res. 2002;36(3):179-187. Statistical analysis: Häuser, Bernardy. 14. Arnold LM, Keck PE Jr, Welge JA. Antidepres- 36. Gonzalez-Viejo MA, Avellanet M, Hernandez- Obtained funding: Häuser, Bernardy, Üçeyler. sant treatment of fibromyalgia: a meta-analysis and Morcuende MI. A comparative study of fibromyalgia Administrative, technical, or material support: Sommer. review. Psychosomatics. 2000;41(2):104-113. treatment: ultrasonography and physiotherapy ver- Study supervision: Häuser, Sommer. 15. Moher D, Cook DJ, Eastwood S, Olkin I, Rennie sus sertraline treatment [in French]. Ann Readapt Med Financial Disclosures: The authors reported that none D, Stroup DF. Improving the quality of reports of meta- Phys. 2005;48(8):610-615. has been employed by pharmaceutical companies and analyses of randomised controlled trials: the QUOROM 37. Gür A, KarakoçM,NasK,CÈevik R, Saraç J, Ataog˘lu that none has stock, options, grants, patents, or pend- statement: Quality of Reporting of Meta-analyses. S. Effects of low power laser and low dose amitripty- ing royalties. Dr Häuser reported receiving honoraria Lancet. 1999;354(9193):1896-1900. line therapy on clinical symptoms and quality of life for educational talks from Lilly, Mundipharma, Pfizer, 16. Bero L, Rennie D. The Cochrane Collaboration: in fibromyalgia: a single-blind, placebo-controlled trial. and Janssen-Cilag. Dr Häuser reported that he will re- preparing, maintaining, and disseminating system- Rheumatol Int. 2002;22(5):188-193. ceive consulting honoraria from Lilly and Pfizer. Dr Ber- atic reviews of the effects of health care. JAMA. 1995; 38. Sencan S, Aki S, Karan A, Müslümanoglu L, Ozcan nardy and Dr Häuser reported receiving travel grants 274(24):1935-1938. E, Berker E. A study to compare the therapeutic effi- for scientific conferences by Lilly and Pfizer. Dr Som- 17. Robinson KA, Dickersin K. Development of a highly cacy of aerobic exercise and paroxetine in fibromyal- mer reported receiving honoraria for educational talks sensitive search strategy for the retrieval of reports of gia syndrome. Back Musculoskeletal Rehabil. 2004; from Boehringer Ingelheim, Genzyme, Pfizer, and controlled trials using PubMed. Int J Epidemiol. 2002; 17(2):57-61. Schwarz-Pharma. 31(1):150-153. 39. Targino RA, Imamura M, Kaziyama HH, et al. A Funding/Support: The study was supported by grants 18. Smythe HA, Moldofsky H. Two contributions to randomized controlled trial of acupuncture added to

Downloaded From: https://jamanetwork.com/ on 09/28/2021 ANTIDEPRESSANT THERAPY IN FIBROMYALGIA SYNDROME

usual treatment for fibromyalgia. J Rehabil Med. 2008; cebo in the treatment of fibromyalgia patients with 59. Patkar AA, Masand PS, Krulewicz S, et al. A ran- 40(7):582-588. or without major depressive disorder. Arthritis Rheum. domized, controlled trial of controlled release parox- 40. Ozerbil O, Okudan N, Gökbel H, Levendog˘luF. 2004;50(9):2974-2984. etine in fibromyalgia. Am J Med. 2007;120(5): Comparison of the effects of two antidepressants on 50. Arnold LM, Rosen A, Pritchett YL, et al. A ran- 448-454. exercise performance of the female patients with domized, double-blind, placebo-controlled trial of du- 60. Vitton O, Gendreau M, Gendreau J, Kranzler J, fibromyalgia. Clin Rheumatol. 2006;25(4):495- loxetine in the treatment of women with fibromyal- Rao SG. A double-blind placebo-controlled trial of 497. gia with or without major depressive disorder. Pain. milnacipran in the treatment of fibromyalgia. 41. Ataog˘luS,Ataog˘luA,Erdog˘ an F, Sarac¸ J. Com- 2005;119(1-3):5-15. Hum Psychopharmacol. 2004;19(suppl 1):S27- parison of paroxetine, amitriptyline in the treatment 51. Carette S, McCain GA, Bell DA, Fam AG. Evalu- S35. of fibromyalgia. Turk J Med Sci. 1997;27:535- ation of amitriptyline in primary fibrositis: a double- 61. Wolfe F, Cathey MA, Hawley DJ. A double- 539. blind, placebo-controlled study. Arthritis Rheum. 1986; blind placebo controlled trial of fluoxetine in 42. CÈapaci K, Hepgüler S. Comparison of the effect 29(5):655-659. fibromyalgia. Scand J Rheumatol. 1994;23(5): of amitriptyline and paroxetine in the treatment of fi- 52. Carette S, Oakson G, Guimont C, Steriade M. Sleep 255-259. bromyalgia syndrome. Pain Clin. 2002;14:223- electroencephalography and the clinical response to 62. Yavuzer G, Küc¸ükdeveci A, Arasli T, Elhan A. Mo- 228. amitriptyline in patients with fibromyalgia. Arthritis clobemid treatment in primary fibromyalgia 43. Carette S, Bell MJ, Reynolds WJ, et al. Compari- Rheum. 1995;38(9):1211-1217. syndrome. Eur J Phys Med Rehabil. 1998;8:35- son of amitriptyline, cyclobenzaprine, and placebo in 53. Ginsberg F, Joos E, Geczy J, Bruhwyler J, 38. the treatment of fibromyalgia: a randomized, double- Vandekerckhove K, Famaey JP. A pilot randomized 63. Kempenaers C, Simenon G, Vander Elst M, et al. blind clinical trial. Arthritis Rheum. 1994;37(1): placebo-controlled study of pirlindole in the treat- Effect of an antidiencephalon immune serum on pain 32-40. ment of primary fibromyalgia. J Musculoskeletal Pain. and sleep in primary fibromyalgia. Neuropsychobiology. 44. Giordano N, Geraci S, Santacroce C, Mattii G, 1998;6:6-17. 1994;30(2-3):66-72. Battisti E, Gennari C. Efficacy and tolerability of par- 54. Ginsberg F, Mancaux A, Joos E, Vanhove P, Famaey 64. Russell IJ, Mease PJ, Smith TR, et al. Efficacy and oxetine in patients with fibromyalgia syndrome: a JP. A randomized placebo-controlled trial of sustained- safety of duloxetine for treatment of fibromyalgia in single-blind study. Cur Ther Res. 1999;60:696- release amitriptyline in primary fibromyalgia. J Mus- patients with or without major depressive disorder: re- 702. culoskeletal Pain. 1996;4:37-47. sults from a 6-month, randomized, double-blind, 45. Goldenberg DL, Felson DT, Dinerman H. A ran- 55. Goldenberg D, Mayskiy M, Mossey C, Ruthazer placebo-controlled, fixed-dose trial. Pain. 2008; domized, controlled trial of amitriptyline and naproxen R, Schmid C. A randomized, double-blind crossover 136(3):432-444. in the treatment of patients with fibromyalgia. Arthri- trial of fluoxetine and amitriptyline in the treatment 65. Living with fibromyalgia, drugs approved to man- tis Rheum. 1986;29(11):1371-1377. of fibromyalgia. Arthritis Rheum. 1996;39(11): age pain. US Food and Drug Administration con- 46. Scudds RA, McCain GA, Rollman GB, Harth M. 1852-1859. sumer updates. http://www.fda.gov/consumer Improvements in pain responsiveness in patients 56. Hannonen P, Malminiemi K, Yli-Kerttula U, Isomeri /updates/fibromyalgia062107.html. Accessed with fibrositis after successful treatment with R, Roponen P. A randomized, double-blind, placebo- December 10, 2008. amitriptyline. J Rheumatol Suppl. 1989;19:98- controlled study of moclobemide and amitriptyline in 66. Higgins JP, Thompson SG, Deeks JJ, Altman DG. 103. the treatment of fibromyalgia in females without psy- Measuring inconsistency in meta-analyses. BMJ. 2003; 47. Anderberg UM, Marteinsdottir I, von Knorring chiatric disorder. Br J Rheumatol. 1998;37(12): 327(7414):557-560. L. Citalopram in patients with fibromyalgia: a ran- 1279-1286. 67. Knapp G, Biggerstaff BJ, Hartung J. Assessing the domized, double-blind, placebo-controlled study. Eur 57. Heymann RE, Helfenstein M, Feldman D. A double- amount of heterogeneity in random-effects J Pain. 2000;4(1):27-35. blind, randomized, controlled study of amitriptyline, meta-analysis. Biom J. 2006;48(2):271-285. 48. Arnold LM, Hess EV, Hudson JI, Welge JA, Berno nortriptyline and placebo in patients with fibromyal- 68. Terrin N, Schmid CH, Lau J, Olkin I. Adjusting for SE, Keck PE Jr. A randomized, placebo-controlled, gia: an analysis of outcome measures. Clin Exp publication bias in the presence of heterogeneity. Stat double-blind, flexible-dose study of fluoxetine in the Rheumatol. 2001;19(6):697-702. Med. 2003;22(13):2113-2126. treatment of women with fibromyalgia. Am J Med. 58. Nørregaard J, Volkmann H, Danneskiold-Samsoe 69. Uceyler N, Häuser W, Sommer C. A systematic 2002;112(3):191-197. B. A randomized controlled trial of citalopram in the review on the effectiveness of treatment with anti- 49. Arnold LM, Lu Y, Crofford LJ, et al. A double- treatment of fibromyalgia. Pain. 1995;61(3):445- depressants in fibromyalgia syndrome. Arthritis Rheum. blind, multicenter trial comparing duloxetine with pla- 449. 2008;59(9):1279-1298.

Downloaded From: https://jamanetwork.com/ on 09/28/2021