Overview
This randomized, multicenter, double-blind, placebo-controlled, parallel-group, phase III trial was conducted in 42 outpatient clinics and hospitals (listed in the Acknowledgments) in Japan between March 2012 and December 2013. An institutional review board for each site (listed in the Appendix) approved the protocol, and all patients provided written informed consent before study commencement. The study was conducted in compliance with the International Conference on Harmonisation Good Clinical Practice guidelines. All monitoring activities for this study were outsourced to SRL Medisearch, Inc., Tokyo, Japan. This study was registered at ClinicalTrials.gov under the identifier NCT01552057 on 9 March 2012. There were no changes to the methods or planned endpoints after study initiation.
Eligibility criteria
The criteria used in a previous study of duloxetine [25] were adopted. Briefly, male and female outpatients aged between 20 and 75 years who met the ACR 1990 criteria for fibromyalgia [2] and had a Brief Pain Inventory (BPI) average pain score ≥4 [26, 27] at visits 1 and 2 were included. Exclusion criteria were as follows: past duloxetine treatment; serious or medically unstable disease, clinically significant abnormal laboratory values, or abnormal electrocardiogram (ECG) findings; pain caused by non-fibromyalgia diseases; poorly controlled thyroid dysfunction; rheumatoid, inflammatory, or infectious arthritis; autoimmune disorders other than thyroid dysfunction; psychiatric disorders other than major depressive disorder within the past year; and suicidal tendencies as assessed using the Columbia-Suicide Severity Rating Scale (C-SSRS) [28].
Patients were prohibited from using analgesics and drugs with analgesic effects, including nonsteroidal anti-inflammatory drugs, anticonvulsants, pregabalin, neurotropin, anesthetics, opioids, and adrenocorticosteroids. The use of analgesics for up to 3 consecutive days and for up to a total of 10 days was permitted only for the treatment of adverse events (AEs). Coadministration of acetaminophen at doses up to 1500 mg/day was permitted to treat AEs and as rescue treatment for fibromyalgia, except on the day before efficacy was evaluated after visit 2 and until just before the evaluation. The use of prophylactic aspirin at doses up to 325 mg/day to prevent cardiac events was also permitted. Patients taking mood-affecting drugs such as antidepressants, sedatives, and benzodiazepines were also excluded. Zopiclone and zolpidem were the only hypnotics permitted during the study, as long as their use began before the participant entered the screening phase (i.e., before visit 1), without dosing changes. Tender point injections and nerve blocks were to be stopped before visit 1. Non-drug therapies (e.g., exercise therapy and cognitive behavioral therapy) received at least 14 days before visit 1 were permitted during the study, as long as no changes were made.
Study design
This randomized, multicenter, double-blind, placebo-controlled, parallel-group phase III trial consisted of four phases. The study lasted 17–18 weeks and included a 1- to 2-week screening phase, a 14-week treatment phase, a 1-week dose-tapering phase, and a 1-week follow-up observation phase (Fig. 1). After the screening phase, patients were assigned randomly to receive duloxetine or placebo in a 1:1 ratio, using a web-based patient registration system (ACRONET Corp., Tokyo, Japan) with a stochastic minimization procedure. The following allocation factors were used: (1) BPI average pain score at visit 2 (<6 vs. ≥6) and (2) presence or absence of concomitant major depressive disorder diagnosed on the basis of the M.I.N.I. International Neuropsychiatric Interview–Japanese version 5.0.0 [29]. It was ensured that the maximum between-group difference in the number of subjects in each medical institution did not exceed two. Blinding was maintained until the end of the study by the person responsible for the study drug assignment.
Duloxetine or placebo was orally administered once daily after breakfast on a double-blind basis. In the duloxetine group, patients received 20 mg for 1 week, followed by 40 mg for 1 week and then 60 mg for 12 weeks during the treatment phase. In the placebo group, subjects received placebo for 14 weeks throughout the treatment phase. Down-titration was performed after completion of the treatment phase or if the patient discontinued after at least 2 weeks of treatment with duloxetine (i.e., if the patient was taking 40 or 60 mg at the time of discontinuation). The drug allocation controller confirmed the study drugs were undiscernible in terms of appearance, packaging, and labeling, and mock titration of placebo pills was also performed to maintain blinding. Only the drug allocation controller was aware of the type of drugs being dispensed.
Outcome measures
Primary efficacy outcomes
The primary efficacy measure was the change in the BPI average pain score from baseline (visit 2) to the end of the 14-week treatment phase (visit 7). The pain score was based on a scale from 0 (no pain) to 10 (pain as bad as patient can imagine).
Secondary efficacy outcomes
Secondary outcomes included the “worst pain severity,” “least pain severity,” and “pain right now” items of the BPI; pain interference with seven daily activities (general activity, walking, work, mood, enjoyment of life, relationships with others, and sleep); data reported in patient diaries; and scores on the Patient Global Impression of Improvement (PGI-I) and the Clinical Global Impressions–Global Improvement (CGI-I) [30]. Pain interference was assessed on a 0 (does not interfere) to 10 (completely interferes) rating scale. Items corresponding to the average and worst pain severity items of the BPI were recorded daily by each patient in the patient diary. The average weekly values were calculated based on the diary. PGI-I and CGI-I [30] were assessed separately by the subjects and physicians on a scale ranging from 1 (very much better/improved) to 7 (very much worse) regarding disease improvement from baseline.
Health outcomes
Patients responded to the Fibromyalgia Impact Questionnaire (FIQ; Japanese version) [31, 32], consisting of 20 questions about the symptoms and discomforts of fibromyalgia. Responses were summed to yield a total score ranging from 0 (no impact) to 100 (maximum impact).
Using the 36-Item Short-Form Survey (SF-36; Japanese version 2) [33, 34], patients assessed their health status by answering 36 questions measuring the following eight subscales, rated 0–100 (with higher scores indicating better health status): physical functioning, physical role functioning, emotional role functioning, general health perceptions, social role functioning, bodily pain, vitality, and mental health.
Using the Beck Depression Inventory II (BDI-II) [35, 36], patients assessed 21 items related to symptoms of depression on a 4-point (0–3) scale. Responses were summed to yield a total score that ranged from 0 to 63 (with higher scores indicating more severe depressive symptoms). The widespread pain index (WPI) and symptom severity (SS) scale of the ACR 2010 criteria [8, 37] were assessed, with maximum scores of 19 and 12, respectively.
Safety outcomes
Safety was assessed on the basis of the presence or absence and incidence of AEs and adverse drug reactions (ADRs) reported during the treatment phase until the end of the follow-up observation phase. Additionally, laboratory tests (hematology, clinical chemistry, and urinalysis), ECG, body weight, and vital signs were measured. The presence or absence of suicidal tendencies was assessed using the C-SSRS.
Statistical analyses
Based on clinical data from three previous studies [20–22], the between-group difference in the change in the BPI average pain score from baseline was estimated to be −0.70 between the duloxetine and placebo groups, with a standard deviation of 2.38 for the change in pain score. Therefore, this study required a total of 370 patients to have a power of at least 80 % at a significance level of 0.05 (two-sided). All efficacy analyses were conducted using the full analysis set (FAS), which comprised all randomized patients who received at least one dose of the allocated study drug and had a baseline and at least one postbaseline BPI average pain score. Safety analyses were conducted using the safety analysis set, which was defined as all randomized patients who were administered the study drugs at least once. Unless otherwise noted, the treatment effects were tested at a two-sided significance level of 0.05.
The primary efficacy measure was analyzed using a mixed-effects model repeated-measures (MMRM) approach to compare the change from baseline in BPI average pain score at week 14 of study treatment between the duloxetine and placebo groups. The model included treatment, week, and a treatment × week interaction as fixed effects, as well as the baseline pain score and presence or absence of concomitant major depressive disorder diagnosed as covariates. The change in BPI average pain score from baseline to week 14 of treatment was also compared between the groups by analysis of covariance (ANCOVA) with the baseline value and presence or absence of major depressive disorder as covariates. The missing data at week 14 of treatment were imputed based on the last observation carried forward (LOCF) approach. Post hoc ANCOVA was used for additional sensitivity analyses of the primary efficacy measure. In these post hoc analyses, the baseline observation carried forward (BOCF) or worst observation carried forward (WOCF) values were carried forward to impute missing data instead of the LOCF method being used. The proportions of responders were calculated for patients with a reduction in the BPI average pain score of ≥30 % from baseline to endpoint or ≥50 % from baseline to endpoint, and also for patients with a sustained response. Sustained response was defined as ≥30 % reduction from baseline to endpoint in the BPI average pain score with a 30 % reduction from baseline at an earlier visit (at least 2 weeks prior) and ≥20 % reduction from baseline for every visit in between if there were any intervening visits. These proportions were compared between the groups using a Mantel–Haenszel test adjusted for the allocation factors.
Unless otherwise noted, the MMRM approach was used to evaluate the differences in the changes in other secondary endpoints from baseline between the duloxetine and placebo groups. PGI-I and CGI-I scores were analyzed using a MMRM model without baseline as a covariate. In post hoc analyses, PGI-I and CGI-I at week 14 were compared between the treatment groups using Wilcoxon’s rank-sum test with the LOCF approach. For SF-36, changes from baseline to week 14 of treatment were compared between the duloxetine and placebo groups using the LOCF ANCOVA approach.
Path analysis, which consists of the following two regression models, was performed to estimate a direct analgesic effect relative to an indirect effect on pain reduction through an improvement in depressive symptoms:
$$ {Y}_1={\alpha}_0+{\alpha}_1{X}_1+{\alpha}_2{Y}_2+{\alpha}_3{Z}_1+{\alpha}_4{Z}_2 $$
and
$$ {Y}_2={\beta}_0+{\beta}_1{X}_1+{\beta}_2{Z}_1+{\beta}_3{Z}_2, $$
where Y
1 is the change from baseline in BPI average pain score, Y
2 is the change from baseline in BDI-II total score, X
1 is treatment, Z
1 is baseline BDI-II total score, and Z
2 is baseline BPI average pain score. The direct and indirect effects of the duloxetine were estimated by α1 and α2 × β1, respectively. Then, the contribution (as a percentage) of each effect to the total effect, which was defined as the sum of the direct and indirect effects, was calculated if feasible.
Incidences of AEs and ADRs were compared between the treatment groups using Fisher’s exact test. The AEs and ADRs reported were coded with MedDRA (version 16.1; Medical Dictionary for Regulatory Activities, McLean, VA, USA), and the incidences were tabulated for each preferred term by treatment group.