Once-daily duloxetine 60 mg in the treatment of major depressive disorder: Multicenter, double-blind, randomized, paroxetine-controlled, non-inferiority trial in China, Korea, Taiwan and Brazil

Abstract

Abstract  The aim of the present paper was to compare the efficacy and safety of duloxetine with paroxetine in the acute treatment of major depressive disorder (MDD). In a randomized, double-blind trial of 8 weeks active treatment, patients with non-psychotic MDD were randomized to duloxetine 60 mg (n = 238) or paroxetine 20 mg (n = 240) once daily. Efficacy was primarily measured on change in the 17-item Hamilton Rating Scale for Depression (HAMD₁₇) using a non-inferiority test with a margin of 2.2. Secondary efficacy measures included the HAMD₁₇ subscales, Hamilton Rating Scale for Anxiety, Clinical Global Impressions–Severity, Patient Global Impressions–Improvement, Somatic Symptoms Inventory and Visual Analog Scales (VAS) for pain. Safety measures included treatment-emergent adverse events (TEAE), vital signs, weight, laboratory analyses and electrocardiograms. Non-inferiority of duloxetine to paroxetine was demonstrated because the upper bound of the confidence interval for mean difference in HAMD₁₇ change (0.71) was less than the non-inferiority margin. Secondary efficacy end-points did not differ significantly between treatments with the exception of VAS back pain, where the pooled mean was lower in the duloxetine group (17.1) compared with the paroxetine group (20.3, P = 0.048). No significant differences were observed in the number of early discontinuations and overall TEAE. However, significantly greater proportions of patients in the duloxetine group experienced nausea and palpitations. No clinically relevant changes in laboratory values, vital signs, weight or electrocardiograms were observed with either treatment. The present study verifies the utility of duloxetine as an efficacious and safe treatment for both emotional and physical symptoms of MDD in this predominantly Asian patient sample.

INTRODUCTION

Major depressive disorder (MDD) is characterized by multidimensional symptoms and, like other serious mental illnesses, leads to impairment of a person's functioning and quality of life, which is often exacerbated by a high level of association with comorbidconditions such as anxiety and alcohol abuse.¹ While emotional aspects of MDD, including depressed mood, anhedonia, feelings of guilt and loss of interest in daily activities, feature prominently in the disease and are recognized as core symptoms in many clinical research rating scales, somatic complaints, particularly painful physical symptoms (e.g. headache, generalized or localized aches and pains), are commonly also present.^1–3 These painful physical symptoms are becoming increasingly recognized as an important component of the illness that needs to be addressed in order to optimize treatment outcomes and decrease the risk of relapse.^4,5

Pioneered by the introduction of tricyclic antidepressants (TCA), the pharmacological treatment of depression is now well established. Although TCA were found to be efficacious, they possess an undesirable side-effect profile, including potentially serious cardiovascular complications.^6,7 Attempts to improve the risk : benefit ratio of antidepressants led to the development of the selective serotonin re-uptake inhibitors (SSRI). While SSRI generally have a more benign side-effect profile, their efficacy is not enhanced over the older TCA^8,9 and there is still room for improvement in the pharmacological treatment of this chronic illness because approximately two-thirds of SSRI-treated patients fail to reach full symptomatic remission.¹⁰ Painful physical symptoms may also be less responsive to SSRI treatment,¹¹ pain predicts longer time to remission,¹² and resolution of physical symptoms is correlated with remission in depressed patients.^5,13 The search for antidepressants with improved efficacy across the breadth of the disease state, which encompasses both emotional and physical symptoms, while simultaneously improving the safety and tolerability profile, has thus continued.

Duloxetine hydrochloride (duloxetine), a serotonin (5-HT) and norepinephrine (NE) re-uptake inhibitor (SNRI), has undergone investigation for the treatment of MDD and has been approved for this indication in the USA, Europe and elsewhere. Due to its relatively balanced 5-HT and NE re-uptake inhibition,¹⁴ it was proposed that duloxetine may not only confer antidepressant effects through its serotonergic and noradrenergic activity, but may also have analgesic properties by acting on descending serotonergic and adrenergic inhibitory pain pathways, leading to improved response and remission of depressive symptoms as a result of greater improvement in the overall symptom spectrum of the disease state.¹⁵ Moreover, because duloxetine is relatively selective for 5-HT and NE transporters, with low affinity for other neurotransmitter receptors,¹⁴ it was predicted to have a favorable side-effect profile.¹⁵

Duloxetine has been studied for the treatment of MDD at daily doses ranging from 40 to 120 mg in a number of acute (8–9-week) randomized, blinded trials with placebo and active control.^15–21 When compared with placebo, these studies have shown duloxetine to be efficacious in the improvement of overall symptoms of depression as measured on the 17-item Hamilton Rating Scale for Depression (HAMD₁₇),^22–23 as well as specific measures of painful physical symptoms.²⁰ However, with some exceptions,²⁴ these studies have been conducted in predominantly Caucasian patient samples either in the USA or Europe. As with any medication, the possibility thus exists that duloxetine will have a different efficacy and/or tolerability profile in other ethnic or cultural settings where differences in the presentation of the disease and clinical practices may exist.²

The objective of the present study was to compare the efficacy and tolerability of duloxetine with paroxetine in a predominantly Asian cohort of patients with MDD. The primary measure of efficacy was the change in HAMD₁₇ total score over 8 weeks of active treatment. The primary analysis, defined a priori, was a non-inferiority test of change in HAMD₁₇ for duloxetine and paroxetine in patients meeting criteria for a defined level of completion and study protocol/drug compliance (i.e. per-protocol analysis). This approach is favored over the intent-to-treat (ITT) approach for testing non-inferiority because, in this situation, it is more conservative and leads to more robust conclusions.²⁵ The non-inferiority margin (delta) chosen was 2.2, in line with the recommendation that delta should be between one-third and one-half of the advantage of the active comparator over placebo and correspond with the minimum difference that would be considered clinically important.

METHODS

Selection of patients

All patients (and/or their authorized legal representative) provided written informed consent prior to the start of any study procedures in accordance with principles of the Declaration of Helsinki. In order to be entered into the study, patients (men and non-pregnant women) must have been at least 18 years of age and met the DSM-IV diagnostic criteria for non-psychotic major depression (single episode or recurrent).³ Baseline severity of symptoms also had to be at least moderate as determined by scores of ≥15 on the HAMD₁₇ and ≥4 on the Clinical Global Impressions–Severity (CGI-S) scale.²⁶ Patients were excluded for the following reasons: current DSM-IV diagnosis other than MDD, previous psychotic disorder diagnosis, dysthymic disorder within the past 2 years, anxiety disorder as a primary diagnosis within the past year, axis II disorder that would interfere with protocol compliance, history of substance abuse, lack of response of the current episode to two or more adequate courses of antidepressant therapy, history of a lack of response to an adequate trial of paroxetine for the treatment of depression, judged to be at serious suicidal risk, serious medical illness, history of hepatic dysfunction, current jaundice, or positive hepatitis B surface antigen (Dane particle; HBsAg) or positive hepatitis C surface antibody (HCAb), alanine aminotransaminase level ≥2-fold the upper limit of normal, electroconvulsive therapy within the past year, psychotherapy, started light therapy or phototherapy within 6 weeks of study entry, taking any excluded medications or abnormal thyroid-stimulating hormone concentrations.

Study design

This study was conducted according to a prespecified protocol (F1J-AA-HMCV) and associated a priori statistical analysis plan. Patients were enrolled at 20 investigational centers in China (seven centers), Korea (six centers), Taiwan (five centers) and Brazil (two centers). Appropriate ethical review boards for each investigational center reviewed and approved the protocol, consent form and any required amendments. First patient visit was 17 February 2004, last patient visit was 12 June 2005.

The study was conducted as a randomized, double-blind, double-dummy, parallel, active treatment-controlled trial with three study periods. Study period I was a 3–14-day screening and washout period consisting of visits 1 and 2. Study period II was an acute 8-week double-blind, active-treatment phase starting at visit 2 (baseline) and consisting of post-baseline visits at 1 week, 2 weeks, 4 weeks, 6 weeks and 8 weeks. In this period, patients meeting the entry criteria were randomized in a ratio of 1:1 to either duloxetine 60 mg/day or paroxetine 20 mg/day. At the investigators' discretion, for patients unable to tolerate the initial dose, the study drug could be halved (i.e. duloxetine 30 mg/day, paroxetine 10 mg/day) at any point in the first 2 weeks of the acute treatment phase, but thereafter the dosage had to be fixed at 60 mg/day for duloxetine and 20 mg/day for paroxetine. Patients unable to tolerate the final fixed dose of study drug were discontinued from the study. Study period III was a dose-tapering period ending with the final visit (visit 8) after 6–10 days and was intended to minimize possible discontinuation symptoms. With the exception of the episodic use of benzodiazepines, concomitant medications with primarily central nervous system activity were not permitted during the study period. Prescription pain medications were not allowed. Anti-hypertensive and other cardiovascular medications were permitted only if the patient had been on a stable dose for at least 3 months prior to the study and remained on the medication for the duration of the study.

Efficacy measures

Efficacy was assessed on the basis of improvement in severity of a broad range of symptoms of depression during the course of acute treatment. The primary measure of efficacy was the HAMD₁₇ total score. Secondary measures of efficacy included HAMD₁₇ subscales–Anxiety/Somatization (items 10, 11, 12, 13, 15, and 17), Retardation (items 1, 7, 8, and 14), Sleep (items 4, 5, and 6), Core (items 1, 2, 3, 7, and 8) and Maier (items 1, 2, 7, 8, 9, and 10), severity of anxiety on the Hamilton Rating Scale for Anxiety (HAMA),²⁷ overall severity of illness on the CGI-S scale, degree of improvement of overall symptoms as assessed by the patient on the Patient Global Impressions-Improvement (PGI-I) scale,²⁶ degree to which physical complaints were bothersome on the Somatic Symptoms Inventory (SSI) scale,²⁸ and patient ratings of physical pain dimensions on a number of visual analog scales (VAS).²⁹ With the exception of the PGI-I (post-baseline only), each of the efficacy measures were recorded at all study period II visits.

Tolerability and safety assessments

The following measures of tolerability and safety were recorded at each acute treatment phase visit: spontaneously reported treatment-emergent adverse events (TEAE), including serious adverse events (SAE), supine systolic and diastolic blood pressure, heart rate, and concomitant medications. Additionally, at screening visits and last acute phase visit or at early discontinuation, weight was recorded and 12-lead electrocardiogram (ECG) and urinalysis conducted. Blood was collected at the screening visit and at 2, 4 and 8 weeks after randomization (or upon early discontinuation) for routine chemistry and hematology laboratory analysis. Abnormal laboratory values were determined by the study sponsor based on established reference limits.³⁰ Treatment-emergent sustained elevation in blood pressure was defined as described elsewhere.³¹

Statistical methods

For the primary efficacy analysis, the non-inferiority of duloxetine 60 mg/day with respect to paroxetine20 mg/day was assessed using the one-sided 97.5% confidence interval of the difference in adjusted baseline to last observation carried forward (LOCF) end-point mean change in HAMD₁₇ total score between the treatments (duloxetine minus paroxetine). Acceptance of non-inferiority was defined a priori as an upper bound of the one-sided 97.5% confidence interval of <2.2 (non-inferiority margin). An analysis of covariance (ancova) model for change in HAMD₁₇ including terms for baseline score, treatment and center was fitted using the per-protocol population to evaluate the one-sided 97.5% confidence interval for treatment. The per-protocol population excluded patients who did not complete 4 weeks of follow-up, were not study drug compliant or had significant protocol violations. The study was designed to provide 90% power to declare non-inferiority of duloxetine with respect to paroxetine. To obtain this level of power, a target of 480 patients was to be randomized (240 per treatment group).

Secondary efficacy analyses of continuous variables were conducted using the HAMD₁₇ total and subscales, HAMA, CGI-S, PGI-I, SSI average score and six VAS pain intensity scores (Overall, Headache, Back, Shoulder, Interference with daily activities, Time in pain while awake). The average treatment effects over the entire study period were assessed using a restricted maximum likelihood (REML)-based, mixed-effects model, repeated-measures approach (MMRM).³² Models included terms for the baseline score, treatment, center, visit, and treatment × visit interaction and baseline score × visit interaction. Significance tests were based on least-squares mean and Type III sum-of-squares for the treatment term, using a two-sided α = 0.05. For comparative purposes, the treatment differences in the adjusted change from baseline to end-point (LOCF) for each measure were evaluated using an ancova model with terms for baseline score, treatment and center.

The rates of response and remission were calculated as secondary measures of efficacy using the LOCF end-point in HAMD₁₇. Response was defined as ≥50% reduction on the HAMD₁₇ total score at end-point. Remission was defined as a HAMD₁₇ total score of ≤7 at end-point. Treatment comparisons in time to event for these outcomes were evaluated using the Kaplan–Meier method and the log–rank test.

For analysis of baseline characteristics and safety variables, treatment comparisons were either made using Fisher's exact test (categorical variables) or an ancova model (continuous variables) that included terms for treatment and investigator where appropriate.

All analyses were conducted with SAS version 8.2 (SAS Institute, Cary NC, USA) and were prespecified in a statistical analysis plan. The primary analysis was based on the per-protocol population defined here. The analysis of TEAE was based on all randomized patients with at least one post-baseline observation and taking at least one dose of their randomized medication (safety set). All other results are based on all randomized patients with at least one post-baseline observation (ITT population).

RESULTS

Patient characteristics and disposition

Of 672 patients screened, 478 were enrolled in the study and randomized to either duloxetine 60 mg/day (n = 238) or paroxetine 20 mg/day (n = 240). Overall, the greatest number of patients were enrolled from China (n = 244), followed by Korea (n = 123), Taiwan (n = 68) and Brazil (n = 43). A summary of overall patient baseline demographic and psychiatric characteristics are presented in Table 1. Of the patients randomized, 69.7% were female, 91.0% were East Asian, and the mean age was 38.5 years. At baseline, the mean weight for patients in the duloxetine group (60.2 kg) was significantly higher than that observed in the paroxetine group (58.3 kg, P = 0.042). There were no other statistically significant differences in demographic characteristics between the two treatment groups. Psychiatric history and severity of illness of the two cohorts were also similar, with no statistically significant between-group differences.

Table 1.  Baseline patient details and psychiatric history by group (all randomized patients) (mean ± SD)

Characteristic	Duloxetine 60 mg/day (n = 238)	Paroxetine 20 mg/day (n = 240)	P
CGI-S, Clinical Global Impressions Severity scale; HAMD17, 17-item Hamilton Rating Scale for Depression; VAS, visual analog scale.
Age (years)	39.0 ± 13.95	38.0 ± 15.27	0.334
Female, n (%)	156 (65.5)	177 (73.8)	0.059
Weight (kg)	60.2 ± 9.99	58.3 ± 10.85	0.042
Ethnicity, n (%)			0.287
East Asian	216 (90.8)	219 (91.3)
Caucasian	17 (7.1)	11 (4.6)
Hispanic	2 (0.8)	4 (1.7)
West Asian	1 (0.4)	5 (2.1)
African	2 (0.8)	4 (1.7)
Psychiatric history
Age at first episode (years)	35.7 ± 13.40	34.5 ± 15.27	0.248
No. previous episodes	2.0 ± 1.30	2.4 ± 2.25	0.086
Duration of current episode (weeks)	36.9 ± 50.12	34.0 ± 63.14	0.525
Symptom profile
HAMD17 total	21.1 ± 4.12	21.2 ± 4.04	0.878
CGI-S	4.4 ± 0.61	4.5 ± 0.65	0.166
VAS pain overall	34.7 ± 27.92	33.6 ± 26.79	0.662

As shown in Fig. 1, the proportion of randomized patients completing the 8-week acute treatment phase was 69.7% in the duloxetine group and 76.3% in the paroxetine group (P = 0.122). The primary reasons for discontinuation were due to patient decision and adverse events. The proportion of patients discontinuing due to ‘patient decision’ was significantly higher in the duloxetine group (n = 42, 17.6%), compared with the paroxetine group (n = 26, 10.8%; P = 0.036). For the primary analysis of non-inferiority of the HAMD₁₇ total score, 189 patients in the duloxetine group and 198 in the paroxetine group met the criteria for inclusion in the per-protocol analysis population. Over the acute treatment phase, compliance with study drug was similar in the two treatment groups. Of all randomized patients, 84.7% of duloxetine-treated patients and 89.0% of paroxetine-treated patients met the definition for study drug compliance (P = 0.216).

Figure 1. Flow chart of patient disposition. There was a significant difference between the treatment groups in the proportion of patients discontinuing due to ‘patient decision’ (P = 0.036).

Primary efficacy analysis

As illustrated in Fig. 2, in the per-protocol population, the upper bound of the one-sided 97.5% confidence interval of the adjusted mean difference in HAMD₁₇ baseline to LOCF end-point change from the ancova model was 0.71 (less than the prespecified non-inferiority limit of 2.2, P < 0.001), thus demonstrating that duloxetine 60 mg/day was non-inferior to paroxetine 20 mg/day in the treatment of MDD based on the primary measure of efficacy.

Figure 2. Schematic representation of non-inferiority test of difference (duloxetine minus paroxetine) in adjusted least square mean baseline to last observation carried forward end-point change in 17-item Hamilton Rating Scale for Depression total score. Because the upper bound of the one-sided 97.5% confidence interval is less than the prespecified non-inferiority margin of 2.2 in both the per-protocol (primary analysis) and intention-to-treat (ITT) populations, non-inferiority is demonstrated. UL, upper limit.

The primary analysis based on the per-protocol population was also supported in the ITT population, in which the upper bound of the one-sided 97.5% confidence interval of the observed adjusted mean difference in HAMD₁₇ change was 1.25, also lower than the prespecified non-inferiority limit of 2.2 (P < 0.001, Fig. 2).

Secondary efficacy analyses

Figure 3 shows the change in HAMD₁₇ from the MMRM analysis. Consistent with the non-inferiority analysis, this indicated no significant advantage for duloxetine 60 mg/day or paroxetine 20 mg/day at any time-point for this measure of efficacy. However, at end-point (8 weeks), the reduction in adjusted mean for the HAMD₁₇ total score in the duloxetine group (−14.19) was numerically greater than in the paroxetine group (−13.52, P = 0.218).

Figure 3. Change from baseline in 17-item Hamilton Rating Scale for Depression (HAMD₁₇) total score: least square (LS) means from mixed-effects model, repeated-measures analysis. Greater reductions indicate more improvement in depressive symptoms. No differences between groups were statistically significant at α = 0.05 at any visit. At 8 weeks, duloxetine = −14.19, paroxetine = −13.52 (P = 0.218).

Table 2 presents the adjusted means from MMRM analysis pooled across all visits. No significant differences were observed between duloxetine 60 mg/day and paroxetine 20 mg/day for the HAMD₁₇ total, HAMD₁₇ subscales, CGI-S, PGI-I, SSI average and five of the six VAS pain measures. In the case of VAS Back Pain, the adjusted mean average of all visits was significantly higher, indicating more severe pain, in the paroxetine group (20.33) than in the duloxetine group (17.14, P = 0.048). Results from a baseline to LOCF end-point change in these variables obtained from ancova analysis showed similar results, although the adjusted mean difference in VAS Back Pain did not reach statistical significance (−2.11, P = 0.290).

Table 2.  Primary and secondary continuous measures of efficacy averaged over all visits (mean ± SE)^†

Measure	Duloxetine 60 mg/day (n = 238)	Paroxetine 20 mg/day (n = 240)	P
^†  Least square adjusted means pooled over all visits from mixed-effects model, repeated-measures analysis of all randomized patients. Adjusted for baseline score, center, visit, treatment-by-visit interaction and baseline score-by-visit interaction. Higher scores indicate greater severity or in the case of PGI-I, less improvement/greater worsening. CGI-S, Clinical Global Impressions Severity scale; HAMA, Hamilton Rating Scale for Anxiety; HAMD17, 17-item Hamilton Rating Scale for Depression; PGI-I, Patient Global Impressions Improvement; SSI, Somatic Symptoms Inventory; VAS, visual analog scale.
HAMD17 total	11.73 ± 0.296	11.94 ± 0.283	0.578
HAMD17 subscales
Anxiety/Somatization	4.35 ± 0.118	4.46 ± 0.112	0.460
Retardation	3.70 ± 0.109	3.81 ± 0.104	0.439
Sleep	2.05 ± 0.082	2.01 ± 0.078	0.703
Core	3.87 ± 0.127	4.03 ± 0.121	0.326
Maier	5.16 ± 0.153	5.34 ± 0.146	0.338
HAMA	11.17 ± 0.294	11.25 ± 0.280	0.837
CGI-S	2.89 ± 0.51	2.95 ± 0.49	0.406
PGI-I	2.75 ± 0.058	2.80 ± 0.055	0.526
SSI average	1.84 ± 0.030	1.87 ± 0.028	0.533
VAS pain subscales
Overall pain	23.31 ± 1.345	25.68 ± 1.284	0.160
Headache	21.29 ± 1.299	21.30 ± 1.240	0.995
Back pain	17.14 ± 1.274	20.33 ± 1.217	0.048
Shoulder pain	18.72 ± 1.226	20.22 ± 1.170	0.332
Interference with daily activities	24.84 ± 1.401	24.61 ± 1.339	0.896
Time in pain while awake	25.64 ± 1.428	26.01 ± 1.366	0.837

For the categorical variables of response and remission, there was no statistically significant difference between treatment groups (Fig. 4). No statistically significant difference was observed between the two groups in terms of time to response and remission. The median number of days to response was 31 for both groups and 46 and 48 days for remission for duloxetine and paroxetine, respectively.

Figure 4. Response and remission at end-point in duloxetine (n = 238) and paroxetine (n = 240) groups. Response was defined as at least 50% reduction on the 17-item Hamilton Rating Scale for Depression (HAMD₁₇) total score at the end-point. Remission was defined as a HAMD₁₇ total score of ≤7 at the end-point. P, Fisher's exact test.

Figure 5 shows the change in the six individual VAS pain measures over time at each visit from MMRM analysis. Consistent with the pooled visit results, with the exception of VAS Back Pain, there were no statistically significant differences between the duloxetine and paroxetine cohorts for these measures of pain at any visit throughout the acute treatment phase. For VAS Back Pain, the greatest separation between groups occurred at 2 weeks after baseline. Indeed, this was the only time point at which the difference between groups in VAS Back Pain reached statistical significance and appears to account for much of the difference averaged over all visits.

Figure 5. Effect of duloxetine 60 mg/day and paroxetine 20 mg/day on pain visual analog scales (VAS); mixed-effects model, repeated-measures analysis of all randomized patients. VAS scales are rated out of 100 by the patient, with higher scores equating to greater pain severity. LS, least square.

Safety

No patient deaths occurred during the present study. Four SAE were reported by patients in the duloxetine treatment group, while none was reported in the paroxetine treatment group. In the investigators' opinion none of the SAE was related to the study drug. Of the four SAE reported, two patients attempted suicide, one patient had pre-existing strabismus, of which the severity increased during the study, and another patient developed viral myocarditis.

Discontinuations due to adverse events (AE) did not differ statistically (P = 0.612) between the two treatment groups, with 20 patients (8.4%) from the duloxetine and 17 patients (7.1%) from the paroxetine group discontinuing from the study as a result of AE. The most frequently reported AE leading to discontinuation in both treatment groups during the acute treatment phase was nausea. The proportion of TEAE (not necessarily leading to discontinuation) reported in the duloxetine treatment group (78.1%) was not significantly greater than that reported in the paroxetine treatment group (70.3%, P = 0.060). Table 3 summarizes the most common TEAE spontaneously reported in the present study. Significantly greater proportions of duloxetine-treated patients experienced nausea and palpitations (37.1% and 9.3%, respectively) when compared with the paroxetine treatment group (24.7% and 4.2%, respectively).

Table 3.  Patients spontaneously reporting TEAE with frequency ≥5% in the duloxetine 60 mg/day group

Adverse Event	Duloxetine 60 mg/day (n = 237) n (%)	Paroxetine 20 mg/day (n = 239) n (%)	P
TEAE, treatment-emergent adverse event.
Nausea	88 (37.1)	59 (24.7)	0.004
Dizziness	50 (21.1)	44 (18.4)	0.491
Dry mouth	41 (17.3)	29 (12.1)	0.122
Constipation	35 (14.8)	27 (11.3)	0.278
Headache	27 (11.4)	29 (12.1)	0.887
Somnolence	27 (11.4)	27 (11.3)	1.000
Palpitations	22 (9.3)	10 (4.2)	0.029
Anorexia	21 (8.9)	17 (7.1)	0.504
Vomiting	19 (8.0)	14 (5.9)	0.373
Decreased appetite	18 (7.6)	19 (7.9)	1.000
Vision blurred	16 (6.8)	16 (6.7)	1.000
Asthenia	13 (5.5)	9 (3.8)	0.392
Fatigue	12 (5.1)	14 (5.9)	0.841
Hyperhidrosis	12 (5.1)	11 (4.6)	0.834

As can be seen from Table 4, there was a slight increase in baseline to LOCF end-point adjusted mean heart rate, and systolic and diastolic blood pressure in both groups, which was not significantly different between groups. Likewise, a small number of patients in both treatment groups experienced sustained elevation in systolic (duloxetine 0.9%, paroxetine 1.7%, P = 0.686) or diastolic (duloxetine 3.2%, paroxetine 0.9%, P = 0.100) blood pressure. A reduction of 0.1 kg in adjusted mean bodyweight was observed in both groups.

Table 4.  Adjusted least square mean baseline to end-point (LOCF) change in weight and vital signs (mean ± SE)

Variable	Duloxetine 60 mg/day (n = 238)		Paroxetine 20 mg/day (n = 240)		P (change)
	Baseline	Change	Baseline	Change
LOCF, last observation carried forward.
Weight (kg)	60.2 ± 0.65	−0.1 ± 0.17	58.3 ± 0.70	−0.1 ± 0.16	0.680
Heart rate (b.p.m.)	77.2 ± 0.57	0.6 ± 0.62	77.4 ± 0.58	0.2 ± 0.61	0.624
Systolic blood pressure (mmHg)	115.6 ± 0.98	1.3 ± 0.78	113.5 ± 0.95	0.8 ± 0.75	0.618
Diastolic blood pressure (mmHg)	75.5 ± 0.65	1.7 ± 0.59	73.7 ± 0.61	1.1 ± 0.57	0.479

While of small magnitude and not considered to be of clinical relevance, statistically significant differences in change from baseline for laboratory values were observed between the two treatment groups in erythrocytes, leukocytes, neutrophils, hematocrit, hemoglobin and urinalysis specific gravity. There were no statistically significant differences between the duloxetine and paroxetine groups in the proportion of patients with treatment-emergent abnormal hematology or hepatic laboratory values. Of the clinical chemistry values assessed, a statistically significant higher proportion of duloxetine-treated patients (3.4%) experienced treatment-emergent abnormal low potassium values, compared with paroxetine-treated patients (0.4%, P = 0.0373) but these instances were isolated, transitory, and not considered clinically significant. No clinically relevant changes in ECG were observed in either of the two treatment groups.

DISCUSSION

This study compared the efficacy and safety of duloxetine, a novel SNRI antidepressant, with paroxetine, a widely used conventional SSRI, in the treatment of MDD. Patients were included from a broad representation of countries/regions (China, Korea, Taiwan and Brazil) outside those traditionally included in antidepressant trials and it is one of only a small number of head-to-head studies comparing duloxetine with an active comparator. It also represents the first randomized, controlled trial of duloxetine for the treatment of MDD in a predominantly Asian sample of patients and thus extends the clinical efficacy and safety profile of this medication to a broader ethnic and cultural context. The 60-mg/day starting and continuation dosage used for duloxetine in the present study has been shown to be the generally optimum therapeutic dose^31,33 and is consistent with the approved label information in other countries, as is the 20-mg/day dosage for paroxetine,^6,34 thus ensuring the clinical relevance of the data reported herein. Moreover, the outcome measures and eligibility criteria applied to patient enrollment in the present study are similar to previous duloxetine clinical trials and the results may therefore be compared with those obtained previously with a reasonable degree of confidence and also pave the way for inclusion of the current results in future meta-analyses.

In the primary analysis of the per-protocol population, duloxetine was found to be non-inferior to paroxetine and this finding was also supported by the ITT population analysis, thus providing a robust confirmation of the non-inferiority of duloxetine to paroxetine on this gold standard measure of efficacy. Results from secondary MMRM analysis of rating scales also indicated that mean scores across all visits during the acute treatment phase did not differ greatly between the groups for measures of depressive symptoms (HAMD₁₇ total and subscales), anxiety (HAMA and HAMD₁₇ anxiety/somatization subscale), global assessments of wellness (CGI-S and PGI-I), and somatic symptoms associated with MDD (SSI average and most of the VAS pain measures). There were no significant differences between groups in the rate of response and remission and the time to these events. However, the MMRM results did show a statistically significant difference between the duloxetine and paroxetine treatment groups for one of the secondary efficacy measures, the VAS Back Pain scale. These results indicated that the mean VAS Back Pain score (across all visits) was significantly greater (more severe level of pain) in the paroxetine treatment group when compared with the duloxetine treatment group. In contrast, this statistically significant difference in VAS Back Pain was not replicated in a more traditional LOCF ancova analysis, although a similar trend was observed (i.e. numerical difference favoring duloxetine).

Although the present study was not placebo controlled, it is relevant to compare our findings to placebo-controlled trials with similar inclusion and exclusion criteria and study duration. In two 9-week acute phase placebo-controlled, predominantly Caucasian patient studies of duloxetine 60 mg/day, Detke et al. observed mean reductions in HAMD₁₇ of −10.9 and −10.5 in the active treatment groups.^15,16 This compared with a mean reduction in HAMD₁₇ total of −14.2 in the present study (MMRM analysis). The level of reduction in the primary measure of efficacy in the present study was thus relatively large, particularly in relation to other studies at doses of 40 to 120 mg/day, which have reported reductions of still smaller magnitude than those reported by Detke and colleagues (summarized in ref. 32). However, the level of response and remission with duloxetine in the present study were comparable to other studies. For example, the rate of remission at end-point in the present study was 49% in the duloxetine cohort, compared with 44% and 43% in other 60-mg/day dosing studies of duloxetine^15–16 and up to 57% in other dosing regimens.³³ The results from the present study also compare favorably to other studies that have included both duloxetine and paroxetine (albeit at different duloxetine doses to the 60 mg/day used here). For example, in pooled results from two studies of duloxetine 80 and 120 mg/day, Perahia et al. found that both doses met statistical criteria for non-inferiority to paroxetine 20 mg/day based on change in the HAMD₁₇ total score.³⁵ As in the present study, these two individual studies (Detke et al. and Perahia et al.) reported similar results for duloxetine and paroxetine on the secondary efficacy measures.^21,35

Although SNRI have been shown to have advantages in terms of tolerability and safety compared with the older TCA,^6,36 previous studies have indicated that there is little, if any, difference between antidepressants, including the TCA and newer classes such as the SSRI and SNRI, in efficacy on traditional emotional symptoms of depression.^8,9,37,38 It is therefore not surprising that in the present study, designed to test non-inferiority and not powered for head-to-head secondary comparisons, that the two compared antidepressants failed to separate on any of the secondary measures of core or overall symptoms of depression because paroxetine was already well established as an efficacious treatment for MDD.^6,9,34 However, it has been proposed that duloxetine (as well as other antidepressants with dual activity), due to its effect on NE and 5-HT re-uptake activity, may have improved efficacy in some painful physical symptoms associated with depression.^15,21,33 When compared with placebo, previous studies have demonstrated the efficacy of duloxetine on a number of pain outcomes in MDD patients and shown that, as with SSRI,¹¹ its antidepressive effects occur largely independently of its analgesic effects.^33,39 In further support of duloxetine's analgesic properties, it has been shown to be effective in non-depressed patients, for example in diabetic neuropathic pain⁴⁰ and fibromyalgia.⁴¹ The results from the current study for the VAS pain measures are therefore noteworthy, with duloxetine having numerical advantages in pooled mean (over all visits) on most VAS pain scales (except ‘interference with daily activities’) when compared with paroxetine. Of particular interest, duloxetine was found to be superior to paroxetine on the VAS Back Pain scale. This difference was most pronounced at around 2 weeks after baseline and accords well with other studies in which duloxetine has consistently separated from placebo on a number of VAS pain end-points, typically at time points starting from around 2 weeks, with VAS Back Pain (along with Overall Pain), usually showing the greatest effect size.^16,39,42

As is typically the case in other duloxetine studies, and studies of antidepressants in general, the most common TEAE seen with duloxetine in the present study was nausea, with dizziness and dry mouth also relatively common as usual.³³ The incidence of nausea with duloxetine (37%) was significantly higher than in the paroxetine group (25%). Interestingly, the incidence of this adverse event is higher in both groups than is often seen with these medications in clinical trials. For example, a pooled analysis of four studies found a rate of 14.4% with duloxetine and 12.0% with paroxetine for nausea.⁴³ However, rates of up to 46% with duloxetine have been reported by Detke et al.¹⁶ It has been suggested that the higher rate of nausea observed in some studies may be due to lack of a dose titration lead-in period,³⁵ and this reasoning may also be applicable to our findings. Nausea is usually mild to moderate and a transitory adverse event associated with duloxetine treatment,⁴⁴ which may partly be alleviated by a temporary dose reduction (allowed in the early part of the present study). Consistent with this, the rate of discontinuation due to nausea was relatively low in the duloxetine group (2.1%), exactly the same as for the paroxetine group. Although occurring at a lower frequency, the incidence of palpitations was also found to be higher in the duloxetine group compared with the paroxetine group in the present study. This is a finding not observed in similar studies previously and given its marginal nature (P = 0.029) and lack of ECG findings, may be due to the multiplicity of comparisons for adverse events giving rise to a spurious result by chance (Type 1 error). However, this finding warrants further attention, particularly in Asian patients, in future studies.

Although the present study provides well-controlled, unbiased and valid data, like any clinical trial, it has limitations that should be taken into consideration when interpreting the results. First, it was essentially a fixed-dose design, conducted under rigid protocol conditions with strict patient eligibility criteria. While these factors serve to enhance the internal validity of the treatment comparisons, they also reduce the degree to which the results can be generalized to usual clinical practice and typical patients who would often not meet eligibility criteria such as lack of comorbidities and concomitant medication use. Moreover, lack of a placebo group limits the conclusions that can be drawn from this trial regarding efficacy in depression and effect size. Also, this was an acute study of MDD and longer term studies are required to better understand the efficacy and safety profile of duloxetine in a chronic condition such as this. Finally, the sample size calculation used to determine enrollment in the present study was based only on the primary objective of non-inferiority in the per-protocol population of patients. While secondary analyses of efficacy and safety have been presented, such head-to-head comparisons should be considered with caution because these are not powered appropriately for such superiority comparisons.

In conclusion, in the present study of predominantly Asian patients, duloxetine (60 mg/day) was shown to be non-inferior to the active SSRI comparator paroxetine (20 mg/day) in the acute treatment of MDD. Both duloxetine and paroxetine appeared to be equally effective in the treatment of MDD-associated symptoms, including anxiety and depression, while greater reductions in back pain were seen in the duloxetine treatment group, when compared to the paroxetine treatment group. In this patient population, both treatments were safe and well tolerated. Thus, the present study verifies the utility of duloxetine as an efficacious and safe treatment for both emotional and physical symptoms of MDD in Asian and Brazilian patients.

ACKNOWLEDGMENTS

The authors thank the clinical investigators, the staff, and the many patients for their participation in this clinical trial. The contribution of the following site investigators is gratefully acknowledged: Chuan Yue Wang, Min-Soo Lee, Xiufeng Xu, Jin Pyo Hong, Niufan Gu, Yong Sik Kim, Jingping Zhao, Doh Kwan Kim, Liang Shu, Woo Taek Jeon, Wenyuan Wu, Jin-Sang Yoon, Ming Li, Chia-Yih Liu, Sandra I Ruschel, Joaquim Mota Neto, Tung-Ping Su, Tzung-Lieh Yeh, Nan-Ying Chiu, Mian-Yoon Chong, The authors would also like to thank Trisha Dwight and Alan Brnabic for helpful comments on the manuscript. This study was sponsored by Eli Lilly and Boehringer Ingelheim. P.L, J.R., S.A.C and G.A.H. are all employees of Eli Lilly. C.Y.W. is a speaker for AstraZeneca, Eli Lilly, GlaxoSmithKline, Janssen, Organon, Watson, and Wyeth and has received research funding from AstraZeneca and Janssen. J.P.H. is on advisory boards for Wyeth Korea and Pfizer Korea and has received honoraria from Eli Lilly Korea. The other authors have no conflicts of interest to disclose.