This publication is provided for historical reference only and the information may be out of date.
We identified 4,850 (1637) citations from searches and reviews of reference lists. We identified an additional 40 citations from dossiers submitted by pharmaceutical companies and 6 from public comments. Some citations were reported in abstract form only and were subsequently excluded (Appendix D).
In all, we included 275 (59) studies: 170 (13) RCTs, 40 (13) meta-analyses, 39 (15) observational studies, and 14 (4) studies of other design. Furthermore, we retrieved 175 (83) articles for background information. Five(Three) studies of interest could not be retrieved after multiple attempts.14–18 Figure 1 (PRISMA flow chart) documents the disposition of the 1067 (278) articles for these studies.
Reasons for exclusions were based on eligibility criteria or methodological criteria (Figure 1, PRISMA flow chart). Seventy-two studies (75 articles) that met the eligibility criteria were later rated as poor quality for internal validity and excluded from the analysis (Appendix C). The two main reasons for a poor quality rating among RCTs were high loss to follow-up (more than 40%) and lack of double-blinding. Among meta-analyses, lack of a systematic literature search was the main reason for exclusions. A lack of systematic literature search leads to a selected spectrum of trials and subsequently to biased results.13
Of 218 (45) included studies, 58 percent were financially supported by pharmaceutical companies; 23 percent were funded by governmental agencies or independent funds. For 19 percent of included studies, we could not determine funding source.
Studies reviewed for this report employed a notable array of diagnostic scales and health status or quality of life instruments. Most were pertinent to depressive and other disorders considered in this report, but some are considered more generic instruments to assess, e.g., health-related quality of life.
Table 4 lists diagnostic scales and health status or quality-of-life instruments encountered in this literature and used in this report.
Key Question 1. For outpatients with depressive, anxiety, adjustment, and/or premenstrual dysphoric disorder, do second-generation antidepressants differ in efficacy?
We included 130 RCTs, 28 meta-analyses, and 1 study of other design. Of the RCTs, 95 were head-to-head trials; 35 were placebo-controlled trials.
I. For adult outpatients with depressive disorder (major depressive disorder and dysthymia subtypes) and pediatric outpatients with major depressive disorder, do second-generation antidepressants differ in efficacy?
A. Major Depressive Disorder in Adults
The following drugs are currently approved by the FDA for the treatment of depressive disorders in adults: citalopram, desvenlafaxine, escitalopram, fluoxetine, paroxetine, sertraline mirtazapine, duloxetine, venlafaxine, bupropion, and nefazodone.
Two comparative effectiveness reviews employing different methods of indirect comparisons of the pharmacological treatment of adult depression have been published.19, 20 Neither review meets formal eligibility criteria because of the inclusion of both in- and outpatients. Nevertheless, we are summarizing the results of both studies because they present the most comprehensive summary of the comparative efficacy and safety of second-generation antidepressants in adult patients with MDD to date.
The first study conducted for AHRQ (Agency for Healthcare Research and Quality) employed head-to-head meta-analyses and indirect statistical methods to evaluate the comparative efficacy for each possible comparison among second-generation antidepressants.19 Authors used meta-regression and network meta-analyses to conduct indirect comparisons of the HAM-D response rates of drugs with insufficient direct head-to-head evidence. They concluded that results from direct and indirect comparisons indicate that no substantial differences exist among second-generation antidepressants. Authors found statistically significant differences for some comparisons, however, the magnitudes of the differential effects were small (less than a relative risk reductions of 15%) and likely not clinically significant.
The second comparative effectiveness review was conducted by the MANGA (Meta-analysis of New Generation Antidepressants) study group.20 Researchers used Baysian- based mixed treatment comparisons to determine the relative effectiveness of drugs that have not been compared in head-to-head trials. Results are different from the AHRQ review. Authors of the MANGA group state that escitalopram and sertraline have the best efficacy–acceptability ratio compared with other second-generation antidepressants. This study however, has been criticized for methodological shortcomings.21–25 Specifically, authors included studies with high risk of bias in their statistical model. In addition, they assumed that a response on the HAM-D scale equals a response on MADRS or CGI despite a lack of evidence to support this assumption.26
Several other meta-analyses confirm that no substantial differences exist between duloxetine and SSRIs,28 escitalopram and SSRIs,29 fluoxetine and SSRIs30 paroxetine and some second-generation antidepressants,31 sertraline and SSRIs,32 venlafaxine and SSRIs,33 and SSRI and SNRI as classes.34
Since the publication of the above mentioned comparative effectiveness reviews, multiple new head-to-head trials have been published.35–52 We have added information on these new studies to Table 6 and describe them in more detail in the chapter on the respective comparisons if they have added new and relevant information.
Fourteen systematic reviews and 75 RCTs compared the effectiveness or efficacy of one second-generation antidepressant to another for treating patients with MDD (Table 6).
Most subjects were younger than 60 years. Inclusion was generally determined on a criteria-based diagnosis (DSM-III-R, DSM-IV]) of MDD and a predefined cut-off point of a universally used depression scale (e.g., HAM-D: 18 or MADRS: 19). Most patients had moderate to severe depression as measured by a variety of scales. Most studies excluded patients who had additional Axis I disorders, high suicidal risk, or progressive medical diseases or who used psychotherapy, electroconvulsive therapy, or psychotropic medications.
Most trials used one or more of the following outcome measures:
- response rate, e.g., more than 50 percent improvement of symptoms on a depression symptoms rating scale, or much or very much improved as assessed by a global assessment method;
- rate of remission; or
- changes in scores on depression scales
Quality of life and functional capacity were rarely assessed, and if they were, they were considered only as a secondary outcome. Most studies employed both physician-rated scales (e.g., HAM-D, MADRS, Clinical Global Impressions Scale [CGI]) and patient-rated scales (e.g., Hospital Anxiety and Depression Rating Scale [HAD-A], Battelle Quality of Life Scale). All studies used physician-rated scales to assess the main outcome measures.
In the majority of studies, the primary endpoints were changes from baseline or rates of response or remission on investigator-rated diagnostic depression scales such as the HAM-D or MADRS. Changes on such diagnostic depression scales are generally viewed as intermediate outcomes rather than health outcomes and are not always reliably related to changes in health outcomes. Response or remission, even when deducted from such a scale (e.g., response is defined as a 50% improvement of scores on HAM-D or MADRS), could be seen as proxies to health outcomes. Therefore, we focused on differences in response or remission rates rather than differences in changes of scores.
Most studies received a fair rating for internal validity. The generalizability of the results was hard to determine and might often be limited. Two European trials53, 54 and one US trial55 in primary care settings, with less stringent eligibility criteria, could be viewed as effectiveness trials. These studies also had long periods of follow-up.54, 55 Drug equivalency was present in all included studies.
Trial reporting was often incomplete. Most articles did not report the method of randomization or allocation concealment. Although last-observation-carried-forward methods (or LOCF analysis, which means that the last observed measurement serves as the substitute for missing values because of the drop out of patients at different time points) were a frequent method of intention-to-treat analysis, few authors reported the overall number of patients lost to follow-up from randomization to the end of the trial. The percentage of imputed measurements, a potential source of bias, was sometimes hard to assess. Many studies did not report the ethnic backgrounds of participants.
Loss to follow-up (number of patients randomized who did not proceed to endpoint), a potential source of bias, was a frequent problem of internal validity. High drop-out rates may be attributable to specific characteristics of a psychiatric outpatient population and a relatively high rate of adverse events in the examined drug class.
1. SSRIs compared to SSRIs in adult outpatients with major depressive disorder
Citalopram compared with escitalopram
Five published trials56–60 and one unpublished61 trial all of fair quality, compared the efficacy of escitalopram and citalopram. Four studies were conducted over 8 weeks, two of them as fixed dose trials56, 57, 59(escitalopram 10 mg/d and 20 mg/d to citalopram 20 mg/d and 40 mg/d). Overall, results favored escitalopram over citalopram. Three studies reported statistically significantly higher response and remission rates for escitalopram than for citalopram. One trial was a fair-rated European/Canadian flexible dose study that compared the efficacy and tolerability of citalopram (20–40 mg/d) to escitalopram (10–20 mg/d) and placebo in 471 depressed outpatients attending primary care centers.56 Loss to follow-up was 7 percent. Intention-to-treat results showed that the escitalopram group had significantly more responders (≥ 50% improvement on MADRS; 63.7% compared with 52.6%; P=0.021) and remitters (MADRS < 12; 52.1% compared with 42.8%; P<0.036) than the citalopram group. Escitalopram was numerically better at all time points on all three efficacy scales (MADRS, CGI-I, CGI-S). The study did not assess health outcomes.
An unpublished, flexible-dose study, derived from the FDA-CDER database, did not find any statistically significant differences in efficacy outcomes between escitalopram and citalopram.61
A pooled analysis of data from three RCTs concluded that escitalopram significantly improved sleep disturbance compared to citalopram.62
It may be significant, however, that both citalopram and escitalopram are produced by the same manufacturer who funded all four available studies. Generic brands of citalopram are available in the US, while escitalopram is still patented.
We conducted two meta-analyses of these studies comparing the effects of citalopram to escitalopram on MADRS scores at weeks 6 to 8. The outcome of the first meta-analysis was the relative risk of being a responder on the MADRS scale (Exhibit 1). A “response” was defined as an improvement of 50 percent or more on the MADRS scale. Pooled results included 1,759 patients and yielded a statistically significant additional treatment effect for escitalopram. The relative risk that a patient would respond was 1.15 (95% CI 1.06 to 1.24) for escitalopram relative to citalopram. Both random effects and fixed effects models presented similar, statistically significant results. The NNT to gain one additional responder based on the pooled risk difference is 13 (95% CI 8 to 39).
The second meta-analysis was an effect size meta-analysis assessing the pooled difference of points on the MADRS scale (Exhibit 2). The weighted mean difference (WMD) presented an additional treatment effect of a 1.52 point reduction (95% CI 0.59 to 2.45; P=0.01) for escitalopram compared to citalopram. Although statistically significant, the clinical significance of the actual difference in effect sizes may be questionable. A 1.3 point change on the MADRS represents about one-fifth to one-quarter of a standard deviation. A recent methods study concluded that, in general, a change of about one-half of a standard deviation on a health-related scale reflects a minimally important difference for a patient.63
Both citalopram and escitalopram are produced by the same manufacturer, which funded all four available studies. Generic brands of citalopram are available in the United States; escitalopram is still under patent protection.
Citalopram compared with fluoxetine
In a fair-rated trial from France, 397 outpatients with MDD attending general practices were randomly assigned to citalopram (20 mg/d) or fluoxetine (20 mg/d) over 8 weeks.64 Loss to follow-up was 12.6 percent. No intention-to-treat analysis was conducted for efficacy measures. Citalopram had a faster onset of efficacy with significantly more patients rated as responding on the MADRS scale (P=0.048) or completely recovered on MADRS and HAM-D scales (P=0.034, P=0.025) after 2 weeks. By 8 weeks, however, MADRS or HAM-D scores showed no statistically significant differences.
Citalopram compared with sertraline
A good-quality Swedish study assessed the effectiveness of citalopram (20–60 mg/d) and sertraline (50–150 mg/d) in 400 patients in general practice during 24 weeks of treatment.53 The majority of patients suffered recurrent depression (sertraline, 56%; citalopram, 65%) and used other medications for medical illnesses (sertraline, 55%; citalopram, 44.5%). Loss to follow-up was 18 percent. The investigators found no significant differences between treatment groups in any measures of depression severity at any point in time (MADRS, Clinical Global Impressions Severity Scale [CGI-S]), Clinical Global Impressions Improvement Scale [CGI-I]). Also, in a subgroup analysis of patients with recurrent depression, they did not report any differences in effectiveness between drugs. Response rates were similar at week 24 (sertraline, 75.5%. citalopram, 81.0%). Treatment groups did not differ significantly in adverse events. This study was one of only a few trials that had not been funded by the pharmaceutical industry.
Escitalopram compared with fluoxetine
A fair, 8-week fixed dose trial evaluated the comparative efficacy of escitalopram (10 mg/d), fluoxetine (20 mg/d), and placebo in depressed patients 65 years or older.65 At study endpoint neither active drug was more efficacious than placebo. MADRS response rates were 46 percent, 37 percent, and 47 percent for patients on escitalopram, fluoxetine, and placebo, respectively. Withdrawal rates were significantly higher among patients on fluoxetine than on escitalopram (17% compared with 26%; P<0.05).
Escitalopram compared with paroxetine
Two fair studies evaluated the comparative effectiveness and safety of escitalopram and paroxetine.43, 44 An 8 -week flexible dose study (escitalopram: 10–20 mg/d; paroxetine 20–40 mg/d) did not identify any statistically significant differences in efficacy between the two treatment groups (MADRS) after 8 weeks of treatment.44 Response (68% compared with 72%) and remission (56% compared with 65%) were similar between patients on escitalopram and paroxetine. The second study, a 24-week fixed- dose trial reported similar findings, however, higher remission rates of patients on escitalopram than on paroxetine reached statistical significance after 24 weeks (75% compared with 67%; P<0.05).43 In both trials patients taking paroxetine had higher discontinuation rates than those on escitalopram. In the fixed dose study, this difference reached statistical significance (32% compared with 19%; P<0.01).43
Escitalopram compared with sertraline
A fair, 8-week trial, funded by the producers of escitalopram, compared fixed-dose escitalopram (10 mg/d) with flexible-dose sertraline (50–200 mg/d) in 212 outpatients with MDD.36 At study endpoint, no differences in efficacy could be detected between the two treatment groups. Seventy-two percent of patients on escitalopram and 69 percent of patients on sertraline achieved HAM-D treatment response, 49% and 53% achieved remission. Other efficacy outcomes (HAM-A, CGI-I, CGI-S, CES-D) were also similar between treatment groups.
Fluoxetine compared with fluvoxamine
Two fair studies evaluated the comparative effectiveness and safety of fluoxetine and fluvoxamine in outpatients with MDD.66, 67 A 7 -week flexible dose study (fluoxetine: 20–80 mg/d; fluvoxamine 100–150 mg/d) did not identify any statistically significant differences in efficacy between the two treatment groups (HAM-D, HAM-A, CGI-S, Raskin-Covi Scale, Hopkins Symptoms Checklist).67 Both treatment regimens significantly improved scores on assessment scales. The second study was a 6-week fixed dose European trial (fluoxetine 20 mg/d; fluvoxamine 100 mg/d) in 184 outpatients with MDD.66 Results are consistent with those of the flexible-dose study; the primary outcome measure (HAM-D) was not significantly different at any time. The drugs were equally effective for secondary outcome measures (CGI, Clinical Anxiety Scale [CAS], the Irritability, Depression, and Anxiety Scale [IDAS], Beck’s Scale for Suicide Ideation [Beck’s SSI]) such as suicidal ideation, sleep, anxiety, and severity of illness at endpoint. Fluvoxamine had significantly more responders on CGI-S (29% compared with 16%; P<0.05) and a greater reduction of CGI-S scores (P<0.05) at week 2 but not at weeks 4 or 6.
Fluoxetine compared with paroxetine
Seven fair-rated studies compared fluoxetine to paroxetine.68–74 Two RCTs were conducted in a population older then 60 years.68, 71 The best trial was an Italian study lasting 1 year that enrolled 242 patients to compare the effects of fluoxetine (20–60 mg/d) and paroxetine (20–40 mg/d) on mood and cognitive function in depressed, nondemented persons (65 years or older).68 Paroxetine had a faster onset of action and a significantly greater improvement of HAM-D scores during the first 6 weeks (week 3: P<0.05; week 6: P<0.002). For up to a year, paroxetine was effective in a higher percentage of patients than fluoxetine (P<0.002 by Kaplan-Meier analysis). Treatment groups did not differ significantly in CGI scores. Fluoxetine had more severe adverse events than paroxetine (22 compared with 9; P<0.002).
The other six studies69–74 lasted 6 to 12 weeks. Loss to follow-up was between 20 and 36 percent. Two studies supported a faster onset of action of paroxetine than fluoxetine,70, 71 four trials did not.69, 72–74 In one study paroxetine-treated patients older than 60 years had a significantly greater response rate on HAM-D and MADRS scales (37.5% compared with 17.5%; P=0.04) than fluoxetine-treated patients. Patients on paroxetine had significantly better Mini Mental State Examination (MMSE) and Sandoz Clinical Assessment Geriatric Scale (SCAG) scores assessing cognitive function at week 3 than did those on fluoxetine. Five studies did not find differences in the improvement of anxiety in patients with depression.68, 69, 72–74 A Canadian RCT assessed anxiolytic activity and akathisia as secondary outcome measures and could not detect any significant differences between treatment groups.69 However, study groups in this trial were not similar at baseline with respect to recurrent depression (paroxetine 76.5% compared with fluoxetine 59.5%), the validity of results might be limited.69
We conducted a meta-analysis of five of these studies (excluding studies that did not report data on HAM-D or were conducted in elderly populations) comparing the effects of fluoxetine to paroxetine on HAM-D scores at the end of follow-up.69, 70, 72–74 A “response” was defined as an improvement of 50 percent or more on the HAM-D scale. The statistical analysis included 690 patients. Results (Exhibit 3) show that the response rate did not differ significantly between fluoxetine and paroxetine (RR: 1.03; 95% CI 0.92 to 1.16) for the random effects model, and the fixed effects model was similarly nonsignificant. Tests for heterogeneity were not significant. Funnel plot, Kendell’s test, and L’Abbe plot did not indicate major biases. However, given the small number of component studies, results of these tests must be viewed cautiously.
Fluoxetine compared with sertraline
Two fair-rated, multicenter trials from France were conducted in office settings (private psychiatrists and general physicians [GPs]).54, 78 The psychiatrists’ study randomized 238 patients for 24 weeks and the GP study 242 patients for nearly 26 weeks (180 days) to fluoxetine (20–60 mg/d) or sertraline (50–150 mg/d). The majority of patients had concomitant medical conditions. Both studies assessed quality of life as a secondary outcome measure (Sickness Impact Profile [SIP], Functional Status Questionnaire [FSQ]). Exclusion criteria were less stringent in the GP trial than the psychiatrist trial. Loss to follow-up was 4.5 percent in the GP trial and 29.8 percent in the psychiatrist trial. In the GP trial, researchers conducted outcome assessments only at day 120 and day 180, but patients could choose to consult the physician at any time. Intention-to-treat analyses in both studies did not reveal any statistically significant differences in any primary (MADRS, HAM-D, CGI) or secondary (Covi Anxiety Scale, HAD, SIP, Leeds Sleep Evaluation) efficacy measures or in the incidence of adverse events.
The ARTIST trial was an open-label RCT designed as an effectiveness study and carried out in a primary care setting (primary care physicians) over 9 months.55 Treatments were randomly allocated. This study enrolled 601 patients at 76 primary care sites. Initial diagnosis for enrollment was not based on diagnostic criteria but rather on the judgment of the treating physician. Criteria-based evaluation classified 74 percent of patients as having MDD, 18 percent dysthymia, and 8 percent minor depression. Patients’ treatments could be switched among study drugs or to other antidepressive medications as needed. Intention-to-treat analysis maintained the original randomization. Outcome measures assessing changes in depression and health-related quality of life measures (work, social and physical functioning, concentration and memory, sexual functioning) were administered over the telephone by a blinded third party. Range of dosage and loss to follow-up were incompletely reported. Results did not reveal any significant differences among drugs in any outcome measures at either 3 or 9 months. All treatment groups significantly improved during the study compared to baseline. Subgroup analyses did not show different effectiveness for patients with MDD or for those older than 60 years.
Three additional fair-rated trials did not find any significant differences in primary outcome measures (HAM-D, MADRS, CGI-S).73, 75, 77, 79 Treatment durations varied from 6 to 16 weeks. One study was conducted in 236 participants older than 60 years.77, 79 In this RCT, outcome measures also included quality of life (Q-LES-Q) and cognitive assessments (Shopping List Task [SLT], MMSE, Digital Symbol Substitution Test). Results on these health outcome measures were similar for both drugs. A subgroup analysis of 75 patients 70 years of age or older showed a greater response rate for sertraline-treated patients (P=0.027).79
We conducted a meta-analysis of five of these studies comparing the effects of fluoxetine to sertraline on HAM-D scores at study endpoint.54, 73, 75–77 All studies except one were financially supported by the manufacturer of sertraline. Results are presented in Exhibit 4. We excluded two studies because different diagnostic scales measured the outcome.55, 78 Our outcome measure was the relative risk of being a responder on HAM-D at study endpoint. A “response” was defined as an improvement of 50 percent or more on the HAM -D scale. Pooled results included 940 patients and yielded a modest additional treatment effect for sertraline just reaching statistical significance. The relative risk of being a responder at study endpoint was 1.13 (95% CI 1.01 to 1.26) for sertraline relative to fluoxetine. Both random effects and fixed effects models presented similar, statistically significant results. The NNT to gain one additional responder based on the pooled risk difference is 13.
A meta-analysis of responders based only on the HAM-D scale did not yield different results. However, all included studies were of fair quality, with some having a loss to follow-up of more than 30 percent. Tests for heterogeneity were not significant. Funnel plot, Kendell’s test and L’Abbe plot did not indicate major biases. However, given the small number of component studies results of these tests must be viewed cautiously.
Paroxetine compared with fluvoxamine
Two RCTs, one flexible-dose80 and one fixed-dose,81 compared the efficacy and safety of paroxetine and fluvoxamine. The flexible-dose trial was a fair 7-week RCT comparing the efficacy and safety of paroxetine (20–50 mg/d) and fluvoxamine (50–150 mg/d) in 60 outpatients with MDD.80 Loss to follow-up was 30 percent. Results presented no statistically significant differences on HAM-D, HAM-A, CGI, and SCL-56. Significantly more paroxetine than fluvoxamine patients suffered from sweating (33% compared with 10%; P=0.028). The fixed-dose trial provided consistent findings.81
Paroxetine compared with sertraline
One fair-rated Swedish RCT compared paroxetine (20–40 mg/d) to sertraline (50–150 mg/d) in a 24-week study.82 A total of 353 patients participated. Outcome measures included MADRS, CGI, and Battelle Quality of Life Measure (BQOL). Loss to follow-up was 35.4 percent. LOCF analysis yielded no significant differences in primary outcome measures (MADRS, CGI) at any point in time. Clinically significant improvement occurred over baseline among all quality-of-life factors. Treatment groups did not differ significantly on BQOL factors. Diarrhea was more frequent in the sertraline group (35.2% compared with 15.2%; P<0.01). Patients in the paroxetine group had higher rates of fatigue (45.8% compared with 21.0%; P<0.01), decreased libido in females (8.8% compared with 1.8%; P<0.05), micturition problems (6.2% compared with 0.6%; P<0.05), and constipation (16.4% compared with 5.7%; P<0.01).
Sertraline compared with fluvoxamine
A fair-rated, 7-week study compared the depression scores and tolerability of sertraline (50–200 mg/d) and fluvoxamine (50–150 mg/d) in 97 depressed patients.83 Loss to follow -up was 30.9 percent. Efficacy did not differ significantly between treatment groups. Both regimens led to significant improvements in depression scores from baseline (HAM-D, CGI). Significantly more patients withdrew because of adverse events in the fluvoxamine group (N=9) than in the sertraline group (N=1; P=0.016). Sertraline-treated patients reported a significantly greater rate of sexual dysfunction (28% compared with 10%; P=0.047).
A fair-rated, small Italian RCT (N=64) randomly assigned asymptomatic patients with a history of unipolar depression and at least one episode within the past 28 months to prophylactic sertraline (100–200 mg/d) or fluvoxamine (200–300 mg/d) treatment for 24 months.84, 85 Patients who remained without recurrence (N=47) prolonged their treatment for another 24 months in an open-label manner. Primary outcome measures were monthly HAM-D assessments. There was no loss to follow-up. Recurrence during the first 2 years of prophylactic treatment did not differ significantly between treatment groups (single recurrence: 21.9% of sertraline-treated patients compared with 18.7% of fluvoxamine patients; z = 0.14, P=0.88). At the 4-year follow-up, no significant differences in recurrences were apparent (sertraline, 13.6%; fluvoxamine, 20%). Adverse events did not differ significantly during the first 24 months of prophylactic treatment.
2. Other second-generation antidepressants compared with SSRIs in adult outpatients with major depressive disorder
Duloxetine compared with fluoxetine
A fair 8-week RCT assigned 173 patients to duloxetine (40–120 mg/d), fluoxetine (20 mg/d), or placebo.86 Overall loss to follow -up was 35 percent. Results revealed no statistically significant differences between duloxetine and fluoxetine in response (49% compared with 45%) and remission (43% compared with 30%). However, the fixed-dose design for fluoxetine but not for duloxetine reduces the validity of this direct comparison.
Duloxetine compared with escitalopram
Three fair, fixed-dose studies compared duloxetine (60 mg/d) to escitalopram (10–20 mg/d).35, 40, 41 The longest study (N=295) lasted 24 weeks.40 An 8-week non-inferiority trial (N=684) did not detect any differences in onset of action or efficacy outcomes (HAM-D) between duloxetine and escitalopram.35 Likewise, after 24 weeks response (73% compared with 77%) and remission (70% compared with 73%) rates were similar between duloxetine and escitalopram. No differences in efficacy could be detected on the HAM-A and CGI-I scales after 24 weeks. In two trials patients on duloxetine had statistically significantly higher discontinuation rates due to adverse events than patients on escitalopram (17% compared with 9%; P<0.05).40, 41
Duloxetine compared with paroxetine
Threefair, 8 -week, fixed-dose trial assessed the comparative efficacy of duloxetine (60 mg/d), duloxetine (80 mg/d), duloxetine (120 mg/d), paroxetine (20 mg/d), and placebo.38, 39, 87 In all three trials efficacy outcomes were similar among duloxetine and paroxetine regimens. In the largest study, 60 percent of patients on duloxetine achieved response and 49 percentremission compared with 65 percent and 50 percent of patients on paroxetine.38 Important to note is that these trials compared a low to medium dose of paroxetine (20 mg) to a medium (80 mg) and high dose (120 mg) of duloxetine.
Mirtazapine compared with fluoxetine
A Taiwanese study compared mirtazapine (30–45 mg/d) to fluoxetine (20–40 mg/d) over 6 weeks in 133 moderately depressed Chinese patients.88 Overall loss to follow-up was 39.4 percent; the drop-out rate was higher in the mirtazapine than the fluoxetine group (45.5% compared with 33.3%; P=NR). LOCF analysis showed no significant differences in any primary outcome measures. More mirtazapine -treated patients than fluoxetine-treated patients reached response and remission at all time points of the study, but none of these differences was statistically significant. No differences in the incidence of adverse events were statistically significant.
Mirtazapine compared with paroxetine
Threetrials assessed the efficacy of mirtazapine (15–45 mg/d) and paroxetine (20–40 mg/d).49, 89, 90 In all three trials, paroxetine and mirtazapine were equally effective in reducing HAM -D and MADRS scores at the endpoint. Mirtazapine led to a faster response in two of the three trials. 89, 90 For example, in a German study, 23.2 percent of mirtazapine-treated patients and 8.9 percent of paroxetine-treated patients responded to the treatment at week 1 (P < 0.002).90 A Kaplan -Meier analysis in the other trial also showed a significantly faster time to response for mirtazapine than for paroxetine (mean 26 days vs. mean 40 days; P = 0.016).89 The NNT to yield one additional patient responding with mirtazapine at weeks 1 or 2 is 7. No significant difference in response rates on the CGI scale was noted. All three trials reported weight gain in significantly more patients treated with mirtazapine than with paroxetine (P < 0.05).
Mirtazapine compared with sertraline
One fair-rated, recent multinational European study examined the onset of efficacy of mirtazapine (30–45 mg/d) compared to that of sertraline (50–150 mg/d) in 346 outpatients.91 Loss to follow-up was 20.8 percent. Onset of action was faster for the mirtazapine group. The mean change of HAM-D scores was significantly greater during the first 2 weeks for mirtazapine than for sertraline (P<0.05); after 2 weeks the difference remained greater but lacked statistical significance. CGI scores did not show significant differences, but MADRS score were significantly greater at week 1 in the mirtazapine group. The Changes in Sexual Functioning Questionnaire did not show significant differences although for mirtazapine the trend was positive. A significantly higher number of patients withdrew because of adverse events in the mirtazapine group (12.5% compared with 3%; P=NR).
Venlafaxine compared with citalopram
A fair European 6-month study compared venlafaxine ER (37.5–150 mg/d) to citalopram (10–30 mg/d) for the treatment of depression in elderly outpatients (mean age 73 years).92 No statistical differences in any outcome measures (MADRS, CGI-S, CGI-I) could be detected at study endpoint. The remission rates were 19 percent for venlafaxine and 23 percent for citalopram. Both treatment groups reached a 93 percent response rate.
Venlafaxine compared with escitalopram
Two fair 8-week studies assessed the comparative effectiveness of venlafaxine XR and escitalopram 93, 94. A fair European, multinational study assigned 293 patients to escitalopram (10–20 mg/d) or venlafaxine XR (75–150 mg/d).93 Results presented no statistically significant differences in response (venlafaxine XR: 79.6%; escitalopram: 77.4%) and remission (venlafaxine XR: 69.7%; escitalopram: 69.9%). Survival analysis of the intention-to-treat population indicated that escitalopram-treated patients achieved sustained remission 6.6 days earlier than patients on venlafaxine XR (P<0.01). Significantly more patients on venlafaxine XR than on escitalopram reported nausea (26% compared with 17%; P<0.05), sweating (12.5% compared with 6%; P<0.05), and constipation (6% compared with 2%; P<0.05).
The second trial reported similar results 94. No statistically significant differences were apparent between venlafaxine XR and escitalopram in response (48% compared with 58.8%) and remission rates. Significantly more patients in the venlafaxine group withdrew because of adverse events (16% compared with 4%; P<0.01) or reported nausea (24% compared with 6%; P<0.05).
Venlafaxine compared with fluoxetine
A South American multicenter study with a good quality rating randomized 382 patients to venlafaxine (75–150 mg/d) or fluoxetine (20–40 mg/d) for 8 weeks.95 Patients were predominantly female and moderately to severely ill. The majority had a previous history of depression (venlafaxine, 79.6%; fluoxetine, 77.4%). Loss to follow-up was 12.3 percent. LOCF analysis yielded no significant differences between study groups in any primary efficacy measures (HAM-D, MADRS, CGI, Hopkins Symptom Checklist). Both treatment groups showed significant decreases of HAM-D and MADRS scores from baseline (P<0.05). Response rates were similar in both treatment groups (venlafaxine, 80.6%; fluoxetine, 83.9%). No significant differences in adverse events were observed.
Three fair-rated studies reported mixed results about the efficacy of venlafaxine and fluoxetine in comorbid patients with high anxiety96, 97 or GAD. 98, 99 Only one study reported significantly greater response rates on HAM-D (71.9% compared with 49.3%; P=0.008) and MADRS (75.0% compared with 49.3%; P=0.001) for venlafaxine than for fluoxetine.96 At the end of the trial, 59.4 percent of venlafaxine-treated patients and 40.3 percent of fluoxetine-treated patients were in remission (P=0.028). All three studies presented greater improvements on anxiety scales (HAM-A, Covi Anxiety Scale) in patients treated with venlafaxine than with fluoxetine. However, differences were only statistically significant in one trial (Covi Anxiety scale: P=0.0004).96 Two studies reported significantly more dizziness (P<0.001) and sweating (P<0.05) in the venlafaxine group than in the fluoxetine group.97–99
Seven additional trials also provided predominantly consistent evidence on a similar efficacy of venlafaxine and fluoxetine.45–48, 100–102 Only one study reported a significantly higher response rate of venlafaxine than fluoxetine (72% compared with 60%; P=0.023).101
We conducted a meta-analysis of eight studies comparing venlafaxine to fluoxetine.45, 47, 96–98, 100–102 All studies were financially supported by the manufacturer of venlafaxine. Three studies were excluded because of missing data.46, 48, 95 The main outcome measure was the response to treatment on HAM-D at study endpoint. Results (Exhibit 5), based on 2,593 patients, show no statistical difference between venlafaxine and fluoxetine (RR 0.04; 95% CI -1.20E-04 – 0.080). Tests for heterogeneity were not significant. Funnel plot, Kendell’s test, and L’Abbe plot did not indicate major biases. However, given the small number of component studies results of these tests must be viewed cautiously.
These findings are similar to results of a meta-analysis recently reported by Smith et al. (2002).103 Venlafaxine showed a modest but statistically significantly greater standardized effect size (−0.14; 95% CI −0.22 to −0.06) and a significantly greater OR for remission (OR 1.42; 95% CI 1.17 to 1.73) compared to fluoxetine. The OR for response was numerically greater for venlafaxine but did not reach statistical significance (OR: 1.17; 95% CI 0.99 to 1.38). This study included inpatients and therefore did not meet the eligibility criteria for this report.
Venlafaxine compared with paroxetine
Two fair studies compared venlafaxine to paroxetine.104, 105 A Spanish study compared venlafaxine (75–150 mg/d) to paroxetine (20–40 mg/d) in outpatients (N=84) with either MDD or dysthymia over 24 weeks.104 The majority (88%) of patients were female. The percentage of dysthymic patients was not reported, and the authors did not differentiate between dysthymia and mild or moderate depression. Loss to follow-up was 32 percent, with a substantially higher loss to follow-up in the venlafaxine group (39% compared with 26%). Intention -to-treat analysis yielded no significant differences between treatment groups on any primary outcome measures (HAM-D, MADRS, CGI) at 24 weeks. However, sample size for this study was small, and it was underpowered because it had been designed as a pilot study.
A 12-week, British fixed-dose trial randomized 361 mainly moderately ill patients (based on CGI severity score) treated in 43 general practices to either venlafaxine XR (75 mg/d) or paroxetine (20 mg/d).105 Loss to follow-up was 27.4 percent. Results revealed no significant differences in efficacy measures, quality of life scores, or adverse events between study groups.
Venlafaxine compared with sertraline
Two good trials106, 107 and one fair37 trial compared the efficacy of sertraline to venlafaxine. A good quality Scandinavian trial compared venlafaxine (75–150 mg/d) to sertraline (50–100 mg/d) in 147 patients who were mainly moderately to markedly ill.107 Study duration was 8 weeks; loss to follow-up was 19 percent. Both treatment groups showed statistically significant reductions in MADRS, HAM-D, and CGI scores. Response rates on the HAM-D scale were higher for venlafaxine at the endpoint (83% compared with 68%; P=0.05), as were remission rates (68% compared with 45%; P=0.008). No significant differences were noted for response or remission rates on MADRS and CGI scales. No significant differences were observed for adverse events. By contrast, the other two studies did not find any differences in efficacy between sertraline(50–150 mg/d) and venlafaxine XR (75–225 mg/d).37, 106
Bupropion compared with SSRIs
A recent, fair-rated meta-analysis compared the benefits and risks of bupropion to SSRIs as a class in 1,332 adult outpatients with MDD.108 The age of the participants ranged from 36 to 70 years. The analysis included five double-blinded, head-to-head RCTs with study durations from 6 to 16 weeks. Three trials assessed the efficacy and safety of bupropion compared with sertraline, one assessed bupropion compared with paroxetine, and one assessed bupropion compared with fluoxetine. The weighted mean differences of CGI-S and HAM-A scores did not differ significantly between bupropion and SSRIs. However, the authors could not pool data on HAM-D and CGI-S because of lack of data.
Bupropion compared with escitalopram
A fair pooled data analysis of two identically designed RCTs assessed the comparative efficacy of bupropion XL (300–450 mg/d), escitalopram (10–20 mg/d), and placebo.42 Both studies lasted 8 weeks and enrolled a total of 830 patients. No differences in efficacy could be detected between the two active treatments (HAM-D, CGI-I, CGI-S, HAD). After 8 weeks, 43 percent of patients on bupropion XL, 45 percent on escitalopram, and 34 percent on placebo achieved remission. Response rates were 62 percent, 65 percent, and 52 percent, respectively.
Bupropion compared with fluoxetine
A fair, 6-week study compared the efficacy of bupropion (225–450 mg/d) and fluoxetine (20–80 mg/d) in 123 patients with moderate to severe depression.109 Loss to follow-up was 27.6 percent but similar in the two treatment groups. Results presented no significant differences in efficacy measures (changes of HAM-D, HAM-A, CGI-S, CGI-I scores). Response rates were similar for both drugs (bupropion, 62.7%; fluoxetine, 58.3%). Adverse events did not differ significantly between treatment groups.
Another fair, 8-week RCT compared efficacy and sexual side effects of bupropion SR (150–400 mg/d), fluoxetine (20–60 mg/d), and placebo in 456 outpatients with MDD.110 Loss to follow-up was 36 percent. Results showed no statistically significant differences in efficacy. At endpoint, bupropion SR had more remitters than fluoxetine (47% compared with 40%). Bupropion SR also showed significantly fewer sexual side effects than fluoxetine throughout the study. Beginning at week 1 until endpoint, significantly more fluoxetine-treated patients than bupropion SR-treated patients (P<0.05) were dissatisfied with their overall sexual function.
Bupropion compared with paroxetine
One fair RCT examined the efficacy of bupropion SR (100–300 mg/d) and paroxetine (10–40 mg/d) in 100 outpatients ages 60 years or older (range 60–88 years) over 6 weeks.111, 112 The majority of patients were white (bupropion SR: 98%, paroxetine: 90%) and female (bupropion SR: 54%, paroxetine: 60%) and had not used antidepressants for the current episode before enrollment (bupropion SR 83%; paroxetine 88%). The overall loss to follow-up was 16 percent with no significant difference between treatment groups. Statistical LOCF analysis showed that efficacy in any outcome measure did not differ significantly between treatment groups. Response rates (≥ 50% reduction in HAM-D scores) were similar in both groups (bupropion SR 71%; paroxetine 77%). Both treatment groups improved significantly in quality-of-life scales (Quality-of-Life in Depression Scale [QLDS], Short Form-36 Health Survey [SF-36]) between baseline and endpoint (P<0.0001), but the treatment groups did not differ significantly.
Bupropion compared with sertraline
A fair, 16-week trial assessed efficacy and tolerability of bupropion SR (100–300 mg/d) and sertraline (50–200 mg/d) in outpatients (N=248) with moderate to severe depression.113 Intention -to-treat analysis with a LOCF method was used to assess main outcome measures. Loss to follow-up was 31.5 percent but similar in the two treatment groups. Efficacy measures (changes of scores on HAM-D, HAM-A, CGI-S, CGI-I) did not differ significantly by treatment group. The article did not report on response or remission rates. Some adverse events (nausea, diarrhea, somnolence, sweating) were significantly higher among sertraline-treated patients (P<0.05). Discontinuation rates because of sexual adverse events were also significantly higher in the sertraline group (13.5% compared with 3.3%, P=0.004).
Two fair-rated RCTs compared the incidence of sexual dysfunction in 360 and 364 patients with MDD during 8 weeks of treatment with bupropion SR (150–400 mg/d), sertraline (50–200 mg/d), or placebo.114, 115 Outcome measures were efficacy (HAM-D, CGI) and sexual dysfunction as assessed by investigators using DSM-IV definitions for sexual dysfunction disorders. Intention-to-treat analyses reported no significant differences in any efficacy measures between bupropion SR and sertraline at endpoints.
During the studies, sertraline showed more sexual adverse events than bupropion at various time points. However, in one trial overall satisfaction with sexual function did not differ significantly between the bupropion and the sertraline group at endpoint.114 In the other study, beginning at day 21 until the end of the study, the overall satisfaction with sexual function was significantly higher in the bupropion group than in the sertraline group (P<0.05).115
Nefazodone compared with fluoxetine
Three studies with identical protocols examined the effects of antidepressive treatment with either nefazodone or fluoxetine on sleep in outpatients with MDD.116–118 Data from these trials were pooled into one analysis.118 A total of 125 patients with MDD and sleep disturbance were enrolled for 8 weeks. Loss to follow-up was 17 percent. Effects on sleep were measured by the Hamilton Depression Rating Scale (HADRS) Sleep Disturbance Factor, Inventory for Depressive Symptomatology-Clinician Related (IDS-C), Inventory for Depressive Symptomatology-Self-Rated (IDS-SR), and EEG measurements.
Nefazodone significantly improved sleep quality as assessed by clinician ratings and self-reported evaluations (P<0.01). Nefazodone and fluoxetine were equally effective in reducing depressive symptoms (changes in HAM-D scores). Response rates for depression were 47 percent for nefazodone and 45 percent for fluoxetine.
Nefazodone compared with paroxetine
Another fair, multi-national study enrolled 206 moderately depressed patients to an 8-week, acute-phase trial comparing nefazodone (200–600 mg/d) to paroxetine (20–40 mg/d).119, 120 Patients who responded to acute treatment were enrolled in an open-label continuation phase (N=108) from week 8 to month 6.120 Overall loss to follow -up was 27.2 percent during the acute trial and 32.4 percent during the continuation phase. Both groups showed significant improvements from baseline HAM-A, HAM-D, and MADRS scores in the acute phase without significant differences between study groups. Clinical improvement was either maintained or improved during the open-label continuation phase without significant differences between groups.
Nefazodone compared with sertraline
A fair, multicenter European study assessed the efficacy and tolerability of nefazodone (100–600 mg/d) and sertraline.121 One hundred-sixty outpatients with moderate to severe depression were enrolled in this 6-week trial. Loss to follow-up was 24.4 percent. Intention-to-treat results did not show significant differences in efficacy between treatment groups. Response rates were similar (nefazodone 59%, sertraline 57%). Additional outcome measures assessed by questionnaire were sexual function and satisfaction under antidepressant treatment. Overall satisfaction with sexual function was significantly higher in the nefazodone group (P<0.01). Among men, 67 percent in the sertraline group and 19 percent in the nefazodone group reported difficulty with ejaculation (P<0.01). Other adverse events did not differ significantly between the two groups.
3. SNRIs compared with SNRIs or other second-generation antidepressants in adult outpatients with major depressive disorder
Venlafaxine compared with duloxetine
The only available head-to-head evidence comparing venlafaxine with duloxetine was a pooled data analysis of two identical RCTs that have not been published individually52. The study pooled results of two RCTs with a 6-week fixed-dose period comparing venlafaxine XR (150mg/d) with duloxetine (60mg/d) followed by a 6-week flexible dose period in 667 patients with MDD. Both RCTs were funded by the makers of duloxetine. Overall, no significant differences in response (69.1 vs. 62.6) and remission (50.3 vs. 48.1) rates could be detected between venlafaxine XR-and duloxetine-treated patients. Discontinuation rates, however, were significantly lower in the venlafaxine than in the duloxetine group (25 percent vs. 35 percent; P = 0.006)
Venlafaxine compared with bupropion
Two 8-week RCTs compared the efficacy and safety of venlafaxine XR and bupropion XR.50, 51 One study was a fixed-dose trial in 591 patients treated with venlafaxine XR (75mg/d), bupropion XR (150 mg/d), or placebo.51 The other study randomized 576 patients to venlafaxine XR (75–150 mg/d), bupropion XR (150–300 mg/d), and placebo.50 After 8 weeks of treatment response, remission rates venlafaxine XR and bupropion XR were similar. For example in the flexible-dose study, MADRS response (65 percent vs. 57 percent; P = NR) and remission rates (51 percent vs. 47 percent; P = NR) did not differ significantly between patients on venlafaxine XR and bupropion XR. Likewise, no substantial differences in health outcomes (Q-LES-Q-SF, Shehan Disability Scale), were apparant at study endpoint.50
4. Summary of the evidence
Seventy-five head-to-head trials compared the effectiveness and efficacy of one SSRI or other second-generation antidepressant to another. All studies addressed initial use of antidepressants. Few studies assessed the efficacy of second-generation antidepressants in comorbid patients with other psychiatric disorders. Patients with other axis I disorders were generally excluded from study participation. Secondary outcome measures often included anxiety scales. Overall, no substantial differences in improvements on anxiety scales exist. However, mixed results or findings limited to a single trial make the body of evidence inconclusive if any of the second-generation antidepressants has a higher efficacy in comorbid patients with high anxiety, recurrent depression, or somatization. A recent systematic review did not detect any differences in efficacy between SSRIs and other second-generation antidepressants for the treatment of MDD with anxiety.122 Generally, high rates of loss to follow-up limit the validity of many studies.
One good53 and two fair-rated54, 55 effectiveness trials provide good to fair evidence that treatment effectiveness does not differ among compared drugs. These comparisons included citalopram to sertraline, fluoxetine to sertraline, and fluoxetine to sertraline and paroxetine. Findings are consistent with evidence from efficacy trials. Two of these trials provide fair evidence that improvement of health-related quality of life (work, social and physical functioning, concentration and memory, sexual functioning) does not differ significantly between fluoxetine, paroxetine, and sertraline.54, 55 The effectiveness of citalopram and sertraline did not differ significantly in a subgroup analysis of patients with recurrent depression.53 However, this finding is limited to a single trial.
Seventy-five efficacy studies and two comparative effectiveness reports conducting indirect comparisons assessed intermediate outcomes such as changes on HAM-D or MADRS scales. Overall, efficacy was similar and the majority of trials did not identify substantial differences among drugs. Statistically significant differences of pooled response rates of some meta-analyses are likely not clinically significant.
Overall discontinuation rates and response and remission rates assessed on multiple diagnostic scales did not differ substantially when taking all the evidence into consideration. We did not find any evidence that one group has a greater benefit from an individual drug than another. Differences among medications exist in speed of response and some aspects of health-related quality of life. For example, mirtazapine presents a faster onset of action than fluoxetine, paroxetine, and sertraline (Table 7); bupropion maintains sexual satisfaction better than escitalopram, fluoxetine, paroxetine, and sertraline (Table 8); and nefazodone improves sleep quality (Table 9).
Three studies yielded fair evidence that mirtazapine has a significantly faster onset of action than fluoxetine, paroxetine and sertraline. The NNT to yield one additional responder at week 1 or 2 is 7. Two additional studies also reported a faster onset of response for mirtazapine than for fluoxetine and paroxetine but differences did not reach statistically significant levels.49, 88 The overall efficacy did not differ significantly between mirtazapine and SSRIs. A well conducted meta-analysis comparing mirtazapine with SSRIs as a class confirms these findings.123 This study did not meet formal eligibility citeria because it included in- and outpatient populations. Nevertheless, findings reported significantly higher response (RR 1.36, 95% CI 1.13 to 1.64) and remission (RR: 1.36; 95% CI 1.20 to 2.36) rates for mirtazapine compared with SSRIs as a class after 2 weeks of treatment.
A pooled analysis of data from three fair-rated trials with identical study protocols comparing nefazodone to fluoxetine reports that improvement of sleep quality is significantly greater in nefazodone-treated patients than in fluoxetine-treated patients.118 All three studies were financially supported by a manufacturer of nefazodone. Similarly, pooled data indicates greater benefits of escitalopram than citalopram in reducing sleep disturbance.62
Muliple studies comparing one SSRI to another provide good to fair evidence that no significant differences exist among SSRIs in improving health-related quality of life or measures of functional capacity (e.g., sleep quality, cognitive function).36, 54, 57, 66, 71, 78, 80–82, 124
Several other efficacy studies assessed quality of life and health-related functional capacity in SSRIs compared to other second-generation antidepressants.37–41, 47, 91, 112, 121 The body of evidence for these comparisons is either inconsistent or based on a single trial. No firm conclusions can be drawn from their results.
B. Dysthymia in Adults
The following drugs are currently approved by the FDA for the treatment of dysthymia in adults: citalopram, escitalopram, fluoxetine, sertraline, mirtazapine, bupropion, and nefazodone.
We did not find any head-to-head trials among patients with dysthymia. Five placebo-controlled studies (Table 10) assessed efficacy and tolerability of fluoxetine, paroxetine, and sertraline in a population with dysthymia.129–136
1. SSRIs compared to placebo in adults with dysthymia
Fluoxetine compared with placebo
A good RCT determined the efficacy and safety of fluoxetine (10–60 mg/d) in elderly patients with dysthymia over 12 weeks.135 ITT results of this NIMH-funded study indicated that fluoxetine had limited efficacy. Response rates on HAM-D did not differ significantly between fluoxetine and placebo (27.3% compared with 19.6%; P=0.4). Likewise, no difference in quality of life could be detected. Statistically significant differences were limited to treatment group – time interactions which presented greater improvements over time on HAM-D and the Cornell Dysthymia Rating Scale (CDRS) for fluoxetine than for placebo.
A second study conducted in patients 18 years or older (mean 43 years) found that fluoxetine had significantly more responders (53.8% compared with 35.9%; P=0.03) than placebo.136 Remission rates favored fluoxetine but did not reach statistical significance (44.4% compared with 25.6%; P=0.07)
Paroxetine compared with placebo compared with behavioral therapy
A large, fair-rated, primary-care-based study randomized 656 patients with dysthymia or minor depression to 11 weeks of paroxetine (10–40 mg/d), placebo, or behavioral therapy.133, 134 Participants were stratified into patients 60 years and older (N=415) and patients younger than 60 years (N=241) for intention-to-treat analysis. Loss to follow-up was not reported for either subgroup.
In the older subgroup, paroxetine-treated patients showed a greater change in Hopkins Symptom Checklist (HSCL-D 20) scores than placebo-treated patients (P=0.004) but not more change than patients on behavioral therapy (P=0.17). For older dysthymia patients with high or intermediate baseline functioning scores, paroxetine improved mental health functioning significantly compared to placebo. Overall, however, improvements for paroxetine-treated dysthymia patients were not statistically significantly different from those on placebo. The younger subgroup did not show statistically significant differences between treatment groups on the HSCL-D scale. For dysthymia only, the remission rate was significantly higher in the paroxetine group than in the placebo group (80% compared with 40%; P=0.008).
Sertraline compared with imipramine compared with placebo
One RCT compared sertraline (50–200 mg/d) to imipramine (50–300 mg/d) and placebo in 416 patients who had had the diagnosis of dysthymia for more than 5 years.129–131 Study duration was 12 weeks; loss to follow-up was 24.3 percent. Outcomes included quality of life and other measures of functional capacity. Both imipramine (64.0%) and sertraline (59.0%) had significantly more responders (CGI 1 or 2) than placebo (44.3%), but the two therapeutic groups did not differ significantly. Quality of life and overall psychosocial functioning improved significantly in both active treatment groups compared to the placebo group. The number of patients who discontinued therapy because of adverse events was significantly higher for imipramine than for sertraline (18.4% compared with 6.0%; P=0.001).
Sertraline compared with placebo
A multinational study enrolled 310 dysthymic patients for 12 weeks to compare sertraline (50–200 mg/d) to placebo.132 Loss to follow-up was 24.2 percent. Patients in the sertraline group had significantly greater reductions in most efficacy measures (MADRS, CGI, HAD-A, HAD-D, Structured Interview Guide for the Hamilton Depression Rating Scale, Seasonal Affective Disorders Version [SIGH -SAD]), than did those in the placebo group. The rates of responders and remitters were also significantly higher in the sertraline group (Hamilton Rating Scale for Anxiety (HAM-A): P=0.001; CGI-I: P<0.001). The quality of life scale (BQLS) showed significantly greater improvements in eight of nine domains in the sertraline group.
2. Summary of the evidence
We identified no head-to head trials. In other trials, significant differences in population characteristics make this evidence insufficient to identify differences between treatments.
One fair study, based in a primary care setting, provides mixed evidence on the effectiveness of paroxetine compared to placebo. A subgroup of patients older than 60 years showed a significantly greater improvement than those on placebo; a subgroup of patients younger than 60 years did not show any difference in effectiveness between paroxetine and placebo.132, 134
Evidence from one good study indicates that fluoxetine has only limited efficacy in elderly patients with dysthymia.135 Fair evidence from two studies indicates that sertraline has a significantly greater efficacy in the treatment of dysthymia than placebo.129–132 In both trials, sertraline treatment led to a significantly greater improvement of quality of life and psychosocial functioning than placebo.
C. Subsyndromal Depressive Disorders in Adults
1. Head-to-head evidence
We did not find any head-to-head RCTs.
Citalopram compared with sertraline
The only head-to-head evidence that we found was a nonrandomized, single-blinded trial (N=138) lasting 1 year which assessed the comparative efficacy and safety of citalopram and sertraline in patients with late-life minor depression or other subsyndromal depressive disorders.137 This study did not meet our formal eligibility criteria. Because it is the only available head-to-head evidence, we are briefly summarizing its results.
Overall, both treatments improved depressive symptoms. No significant differences in efficacy could be detected at any time point. At the end of the study, remission was achieved by 53 percent of patients on citalopram and 42 percent on sertraline (P=0.25). Likewise, no differences in psychosocial functioning emerged.
2. Placebo-controlled evidence
Two studies were conducted in populations with minor depression.
Fluoxetine compared with placebo
A 12-week trial (N = 162) evaluated the efficacy of fluoxetine in patients with minor depression.138 Improvements on depression scales (HAM-D, Beck Depression Inventory [BDI], IDS-C) were statistically significantly greater for patients receiving fluoxetine than for those receiving placebo. Likewise, the overall severity of illness (CGI-S) improved statistically significantly more in the fluoxetine than in the placebo group (P=0.002). No significant differences could be detected in psychosocial outcomes.
Paroxetine compared with placebo
A large primary-care-based effectiveness study randomized 656 patients with dysthymia or minor depression to 11 weeks of paroxetine (10–40 mg/day), placebo, or behavioral therapy.133, 134 Participants were stratified into patients 60 years and older (N=415) and patients younger than 60 years (N=241) for ITT analysis.
In the 60 or older subgroup, patients receiving paroxetine showed a greater change in HSCL-D-20 scores than those receiving placebo (P=0.004), but those on paroxetine did not demonstrate more change than patients on behavioral therapy (P=0.17).134 Effects were similar for patients with dysthymia and minor depression. Paroxetine was not more efficacious than placebo in patients with minor depression in the younger subgroup.133
3. Summary of the evidence
We identified no head-to head RCT. The only available head-to-head evidence was a nonrandomized, open-label trial comparing citalopram with sertraline.
A nonrandomized open-label trial did not detect any differences in efficacy between citalopram and sertraline.137 In placebo-controlled trials, significant differences in population characteristics make the evidence insufficient to identify differences between treatments.133, 134, 138
D. Seasonal Affective Disorder in Adults
Currently, only bupropion has FDA-approval for the treatment of seasonal affective disorder. As in other chapters, we view FDA-approval as evidence for general efficacy, and therefore do not review placebo-controlled trials on drugs that have been FDA-approved.
We found three publications that met our eligibility criteria. These describe two studies assessing SSRIs, one placebo controlled trial of sertraline, and one head-to-head RCT comparing fluoxetine to light therapy.139–141 We excluded two studies because they had a study duration of 5 weeks, which did not meet our eligibility criteria. Nevertheless, we briefly summarize them in the following paragraphs due to lack of evidence for this indication.142, 143 No second-generation antidepressants were compared to one another.
Inclusion of patients was determined by a criteria-based (DSM-III-R, DSM-IV) diagnosis of major depressive episodes with a seasonal pattern,140 or more broadly, major depression, depressive disorder NOS, bipolar disorder depressed, or bipolar disorder NOS with a seasonal pattern.139 Both studies also used seasonal affective disorder specific evaluation tools, either the Hamilton depression scale HAM-D-24, consisting of the HAMD-17 plus 7 addition seasonal affective disorder specific criteria, or the SIGH-SAD (Structured Interview Guide for the Hamilton Depression Rating Scale, Seasonal Affective Disorders Version). In addition, all patients were enrolled during winter.
1. SSRIs compared to placebo in adult outpatients with seasonal affective disorder
Sertraline compared with placebo
One fair study randomized 187 outpatients with DSM-III-R criteria for either major depression, depressive disorder NOS, bipolar disorder depressed or bipolar disorder NOS with a seasonal pattern to 8 weeks of sertraline (50–200 mg/d) or placebo.139 Sertraline was better than placebo at endpoint in the ITT population for all of the outcomes measured, including both physician (HAM-D-29, HAMD-21, HAM-D-17, HAM-D item 1, CGI-S, HAM-A) and patient assessed (HAD-D, HAD-A) measures of depression and anxiety. 62.4 percent of patients in the sertraline group achieved a CGI-I response (rating of one or two), compared with 46.2 percent in the placebo group, P=0.04. The mean final dose of sertraline was 111.3 ± 44.9 mg/d.
Fluoxetine compared with placebo
One fair study randomized 68 patients to treatment with either fluoxetine (20 mg/d) or placebo.143 The study duration of 5 weeks did not meet our eligibility criteria, however we mention it here due to lack of evidence. Clinical response, defined as a greater than 50 percent reduction in HAM-D-29 over the five weeks, was achieved by 59 percent of the fluoxetine group compared to 34 percent of the placebo group, a statistically significant result (P<0.05).
2. SSRIs compared to light therapy in adult outpatients with Seasonal Affective Disorder
Fluoxetine compared with light therapy
One good RCT compared fluoxetine 20 mg/d to light therapy (10 000 lux, 30 minutes/day between 7:00am and 8:00 am) in 96 patients with DSM-IV criteria for major depressive episodes with a seasonal pattern over 8 weeks.140 Primary outcomes measured were clinical response and remission, based on a reduction in HAM-D-24 of greater than fifty percent (response), plus a score of eight or less at endpoint (remission). Both fluoxetine and light therapy were shown to be effective over time, but there were no differences in clinical response rate (both 67%) or remission (54% and 50%, respectively). A subgroup analysis of severely depressed patients, defined as a HAM-D-24 of at least 30, also revealed comparable response (73% compared with 70%) and remission (50% compared with 48%) rates.
An additional fair RCT comparing 5 weeks of fluoxetine 20 mg/d to light therapy (3000 lux, 2h/d, morning or evening) in 40 patients did not meet our eligibility criteria because of its short duration.142 Results, however, were consistent with findings reported in the trial above. Seventy percent of patients treated with light therapy and 65 percent of the fluoxetine group achieved a response to treatment. Numerically more patients on light therapy than on fluoxetine achieved remission (50% compared with 25%; P=0.10)
3. Summary of the Evidence
No head-to-head evidence was available. We identified two trials, one comparing sertraline to placebo, and one comparing fluoxetine to light therapy.
We did not identify any study with a high degree of generalizability.
E. Major Depressive Disorder in Children and Adolescents
Currently, fluoxetine is the only second-generation antidepressant approved by the FDA for treating MDD in both children (2 to 12 years) and adolescents (13 to 18 years). Based on two RCTs,144, 145 escitalopram was approved in 2009 for the acute and long-term treatment of adolescents (12 to 18 years) suffering from MDD. Published evidence is based on controlled clinical trials of children and adolescents 7 to 18 years of age. Fluvoxamine and sertraline are approved for the treatment of OCD in pediatric patients, although they are not approved for treating MDD.
In September 2004, the FDA completed a review of existing data for the risk of both suicidal ideation and suicide attempts in children taking antidepressant drugs for MDD. Based on this review, the FDA instructed the manufacturers of all antidepressants included in this review to revise the labeling for their products to include a boxed warning and expanded warning statements that alert health care providers to an increased risk of suicidality (suicidal thinking and behavior) in children and adolescents being treated with these agents. The FDA’s analysis was based on pooled data from short-term (4 to 16 weeks) placebo-controlled trials of 9 antidepressant drugs (SSRIs and others). This analysis revealed a greater risk of adverse events representing suicidal thinking or behavior (suicidality) during the first few months of treatment in those receiving antidepressants. Although no suicides occurred in these trials, the average risk of such events was 4 percent in patients taking antidepressants—twice the placebo risk of 2 percent.
In the published literature, we did not identify any head-to-head trials comparing one second-generation antidepressant to another for treating MDD in children and adolescents. We found seven fair controlled trials comparing a non-FDA-approved SSRI or SNRI to placebo (Table 13). Additionally, one good-rated trial compared fluoxetine, cognitive-behavioral therapy (CBT), and fluoxetine plus CBT to placebo.
In addition, three systematic reviews evaluated placebo-controlled evidence for the use of SSRIs and an SNRI.146–148 Two reviews highlighted placebo-controlled evidence already included in this discussion,147, 148 so we do not comment on them further here. One review, however analyzed published and unpublished data for citalopram, fluoxetine, paroxetine, sertraline, and venlafaxine.146 We cite the evidence reported in this article because of its contrast with other published evidence.
Of the primary studies evaluated, patient populations generally were between the ages of 6 and 18 years. In general, inclusion was determined by a combination of several factors, often including a criteria-based diagnosis for MDD (DSM-III, DSM-IV) in addition to a predefined severity of disease (HAM-D ≥ 12; CDRS-R > 40; Children’s Global Assessment Scale < 60, Montgomery-Åsberg Depression Rating Scale [MADRS] ≥ 16). Several studies used different inclusion cut-off points when defining severity of disease. All studies lasted between 6 and 12 weeks. Patients were excluded if they were suicidal, had a current or past failure on a study drug, had a seizure disorder, or had a current or past history of bipolar disorder, panic disorder, schizoaffective disorder, OCD, or other significant mental illness.
Primary outcome measures included mean change in score on a standardized depression rating scale (Children’s Depression Rating Scale Revised [CDRS-R]), HAM-D, MADRS, or the Children’s Depression Inventory [CDI]), response (≥ 40%–50% reduction in depression score), or remission (≤ 8 on the HAM-D). Secondary efficacy measures included additional measures of improvement, depression, or anxiety (CGI-I, 9-item subscale of the Kiddie Schedule for Affective Disorders and Schizophrenia for Adolescents-Lifetime version [K-SADS-L], MADRS, HAM-A, Mood and Feelings Questionnaire [MFQ]), and multiple domains of functioning, general health, behavior, and quality of life (Autonomous Function Checklist for parents, Self-Perception Profile, Sickness Impact Profile, Global Assessment of Functioning [GAF] Scale, Child Behavior Checklist [CBCL], Children’s Global Assessment Scale [CGAS], Pediatric Quality of Life Enjoyment and Satisfaction Questionnaire [PQ-LES-Q]).
1. SSRIs compared to placebo in pediatric outpatients with major depressive disorder
Citalopram compared with placebo
One 8-week study randomized 174 children (7 to11 years) and adolescents (12 to 17 years) with MDD to citalopram (20–40 mg/d) or placebo.149 Diagnosis was established with the Schedule for Affective Disorders and Schizophrenia for School-Age Children – Present and Lifetime Version (K-SADS-PL). Overall loss to follow-up was 22 percent. The primary outcome was the mean change from baseline to endpoint in the CDRS-R. Secondary outcome measures included the CGI-I and CGI-S. At 8 weeks, intention-to-treat analysis confirmed significantly greater reduction in the CDRS-R for citalopram-treated patients then for placebo-treated patients (P<0.05). Significant differences were not reported for secondary outcome measures. More than 10 percent of citalopram-treated patients experienced rhinitis, nausea, and abdominal pain (P=NR for comparison with placebo).
Fluoxetine compared with placebo
Although we did not review placebo-controlled evidence for fluoxetine because the FDA has already established its general efficacy and tolerability, we did review the Treatment for Adolescents with Depression Study (TADS) because it specifically compared fluoxetine, fluoxetine plus CBT, CBT alone, and placebo.150 In this good, 12-week, US-based multicenter study of 439 adolescents (12 to 17 years), placebo and flexible-dose fluoxetine (10–40 mg/d) were administered double-blind; CBT alone and CBT with fluoxetine were administered unblinded. Primary outcome measures included the CDRS-R and CGI-I. Overall loss to follow-up was 18 percent. Compared to fluoxetine alone (P=0.02) and CBT alone (P=0.01), treatment with fluoxetine plus CBT was superior on the CDRS-R. Both fluoxetine alone (P<0.001) and fluoxetine plus CBT (P<0.001) demonstrated significantly greater improvement on the CGI-I compared to placebo. Differences in harm-related adverse events were not significant across treatment groups (P=0.15). The trial was subsequently extended to 36 weeks in an open label manner.151 327 patients completed the trial, which did not include a placebo arm, and demonstrated equivalent effectiveness between fluoxetine, CBT and combination therapy (response rates 81% compared with 81% compared with 86%, respectively). Suicidal events were more common in the fluoxetine only group compared to the CBT only and combination groups across the 36 weeks of treatment (14.7% compared with 6.3% compared with 8.4%, respectively). Ten percent of the patients included in the TADS study reported at least one event related to suicidality. 152
Paroxetine compared with placebo
Three multicenter, double-blinded, randomized-controlled trials compared flexible-dose paroxetine to placebo.153–155 One 8-week study conducted in 12 centers in the US and Canada randomized 275 adolescents (12 to 18 years) to double-blind treatment with paroxetine (20–40 mg/d), imipramine (200–300 mg/d), or placebo.153 One fair international study based in South Africa randomized 286 patients aged 13–18 to 12 weeks of paroxetine 20–40 mg/day or placebo,154 and one fair US based trial randomized 206 patients aged 7–17 to 8 weeks of paroxetine 10–50 mg/day or placebo.155 All patients met DSM-IV criteria for MDD. Patients were generally excluded if they had another psychiatric condition or posed a serious suicide risk. The primary outcomes were HAM-D, CDRS-R, MADRS and K-SADS-L depression subscale score. Secondary measures included CGI-I, CGI-S, BDI, MFQ.
All three studies reported similar response rates between patients treated with paroxetine and placebo. For example in the South African study, in 13–18 year old patients a reduction in MADRS of greater than 50 percent was achieved in 60.5 percent of the paroxetine group and 58.2 percent of the placebo group.154 A post hoc sub-group analysis of patients 16 or younger demonstrated a numerical advantage for placebo over paroxetine in MADRS response (placebo 64.9% compared with paroxetine 55.1%). Similarly, the US study of 7–17 year olds demonstrated no difference between paroxetine and placebo in any outcome (change in CDRS score, CGI-I or CGI-S). The post hoc sub-group analysis of 7–11 year old children also revealed a trend for better outcome with placebo over paroxetine (change in CDRS 5.3 points in favor of placebo, P=0.054). Vomiting, dizziness, sweating and suicide-related adverse events were more frequent in the paroxetine group.
Sertraline compared with placebo
One published multinational (US, India, Canada, Costa Rica, and Mexico) study pooled data from two double-blind RCTs conducted in 53 centers.156 These identically designed, concurrently conducted 10-week trials randomized 376 children and adolescents (6 to 17 years) to flexible-dose sertraline (50–200 mg/d) or placebo. Significantly more sertraline-treated patients were female (P=0.02). Twenty percent of randomized participants did not complete the study. The primary efficacy measure was mean change from baseline score on the CDRS-R. In the intention-to-treat analysis, sertraline-treated patients had a significantly greater mean change in CDRS-R score (P<0.01). Significant differences were observed as early as week 3. Secondary efficacy measures included treatment response (≥ 40% decrease in CDRS-R or CGI-I score of 2 or lower), symptoms of anxiety (Multidimensional Anxiety Scale for Children [MASC]), patient’s social functioning [CGAS], and quality of life [PQ-LES-Q]). Significantly more sertraline-treated patients were defined as treatment responders (P<0.05). Statistically significant differences were not observed for measures of anxiety, social functioning, or quality of life. Sertraline-treated patients reported a higher incidence of insomnia, diarrhea, vomiting, anorexia, and agitation.
Of note for this study is the fact that only pooled data from the two independent trials were published. Before this pooling, neither trial had demonstrated a consistent advantage for sertraline over placebo (data available at http://medicines.mhra.gov.uk). One trial reported significantly more sertraline-treated CDRS-R responders (P=0.033 compared to placebo).
2. SNRIs compared to placebo in pediatric outpatients with major depressive disorder
Venlafaxine compared with placebo
One 6-week trial randomized 40 children and adolescents (8 to 18 years) to treatment with venlafaxine and psychotherapy or placebo and psychotherapy.157 Of participants randomized to active treatment, children (8 to 12 years) received venlafaxine in fixed doses of 37.5 mg/d and adolescents (13 to 18 years) received fixed doses of 75 mg/d. An intention-to-treat analysis was not conducted, thereby excluding 17.5 percent of participants randomized to venlafaxine or placebo (15% and 20%, respectively). Efficacy measures evaluated mean change from baseline on two clinician-rated depression scales (HAM-D and CDRS-R), a patient-rated symptoms scale (CDI), and a parent-rated measure of behavioral functioning (CBCL). Compared to placebo, statistically significant differences from baseline were not reported for any of the efficacy measures. A higher percentage of patients experienced side effects in the venlafaxine group than in the placebo group at almost every treatment week.
3. Systematic reviews of published and unpublished data comparing SSRIs and SNRIs to placebo in pediatric outpatients with major depressive disorder
Three systematic reviews evaluated published and unpublished studies comparing a SSRI or SNRI to placebo in children and adolescents.146–148 The largest report reviewed placebo-controlled studies on citalopram, fluoxetine, paroxetine, sertraline, and venlafaxine, including data for 2,145 randomized participants (5 to 18 years).146 The authors abstracted data on remission and response (where appropriate criteria were used), and mean depression score. Scales and responder definitions were different for each study. Risks were assessed by abstracting data on suicide-related behaviors and discontinuation of treatment due to adverse events. Risk-benefit profiles were evaluated for each drug. Fluoxetine was the only second-generation reported to have a favorable risk-benefit profile. Data from two unpublished citalopram trials supported a negative risk-benefit profile, although evidence of efficacy was stated to be limited. Published and unpublished data combined for paroxetine demonstrated no improvement in depressive symptoms and little effect on response; additionally, an increased risk of serious adverse events was reported. Unpublished data on sertraline indicated that it may be even less effective than reported in published trials. Combined, published and unpublished data on venlafaxine suggested a negative risk-benefit profile.
This review highlights distinctions between published and unpublished studies, revealing the potential for publication bias. In this study that reviewed more comprehensive evidence than published studies alone, the authors concluded that fluoxetine is the only second-generation antidepressant to demonstrate a favorable risk-benefit profile for the treatment of pediatric outpatients with MDD.
4. Summary of the evidence
We did not identify any head-to-head trials. Published evidence is insufficient to compare one second-generation antidepressant to another in pediatric outpatients with MDD.
We did not identify any study with a high degree of generalizability.
The existing evidence, summarized in three systematic reviews of published and unpublished RCTs, provides fair evidence that efficacy to improve health outcomes does not differ between placebo and citalopram, sertraline, paroxetine, and venlafaxine.146–148 These studies support a greater efficacy for fluoxetine compared to placebo. No evidence exists for duloxetine, fluvoxamine, mirtazapine, bupropion, and nefazodone.
II. For adult outpatients with anxiety disorders (generalized anxiety disorder, obsessive compulsive disorder, panic disorder, post-traumatic stress disorder, social anxiety disorder), do second-generation antidepressants differ in efficacy?
A. Generalized Anxiety Disorder
Currently, two SSRIs – escitalopram and paroxetine – are approved by the FDA for the treatment of GAD (generalized anxiety disorder). In addition, one SNRI (venlafaxine) and one SSNRI (duloxetine), are approved for the treatment of GAD.
Four head-to-head trials compared one second-generation antidepressant to another for the treatment of GAD.158–161 Two are rated fair158, 160 and two rated poor.159, 161 Additionally, we identified two trials (three publications) that assessed efficacy and tolerability of sertraline,162–164 an SSRI currently not FDA-approved for GAD.
Across reviewed studies that assessed health outcomes, the populations examined were 18 to 80 years of age. Inclusion was based on a criteria-based diagnosis (DSM-IV) of GAD with a minimum score of 18 or 20 on the HAM-A and a score of two or higher on the anxious mood and tension items of the HAM-A. Patients were excluded if they were considered to have MDD, generally defined by a score of 16–17 or higher on the MADRS.
1. SSRIs compared to SSRIs in adult outpatients with GAD
Escitalopram compared with paroxetine
A fair rated RCT compared escitalopram to paroxetine (and placebo) in 681 patients over a 12 week duration.158 All active arms were found to improve the symptoms of GAD compared to placebo. Escitalopram 10 mg was shown to be more effective than paroxetine 20 mg. In the case of CGI-I, escitalopram 10 mg was significantly superior to paroxetine 20 mg at week 12, P<0.05 (Data = NR) and the difference in the HAM-A at 12 weeks was −2.06 (95% CI −3.90 to −0.21, P<0.05).
Paroxetine compared with sertraline
One fair rated small RCT compared paroxetine (10–40 mg/d) to sertraline (25–100 mg/d) in 55 patients with GAD.160 Study duration was 8 weeks. At study endpoint no statistically significant differences in any outcome measures were apparent. Both treatment groups experienced significant reductions in HAM-A scores with similar response (paroxetine 68%, sertraline 61%) and remission rates (paroxetine 40%, sertraline 46%). Likewise no differences could be detected in quality of life outcome measures.
2. SSRIs compared to SNRIs in adult outpatients with GAD
Escitalopram compared with venlafaxine XR
One fair rated RCT (n = 404) compared escitalopram to venlafaxine XR (and placebo) over an 8 week duration.165 The least square mean difference for venlafaxine XR and for escitalopram was similar (P = not reported).
In the case of CGI-I the response rates were also similar between escitalopram (60%) and venlafaxine XR (65.6%). Discontinuation rates due to adverse events were higher for venlafaxine XR (13%) than for escitalopram (7%), but the P-value was not reported.
Paroxetine compared with venlafaxine
A poor quality study compared venlafaxine and paroxetine.159 This small study with 46 participants and a high drop-out rate of 30 percent found no difference between the two treatments. The rates of response (≥ 50% reduction in the HAM-A) were 90.5 percent for venlafaxine compared with 92 percent for paroxetine (P=0.855).
3. SNRIs compared to SSNRI in adult outpatients with GAD
Venlafaxine compared with duloxetine
A fair rated (n=581) RCT 166, which compared duloxetine 20 mg, duoloxetine 60–120mg and venlafaxine XR 75–225mg found no differences among the treatments. In this 10-week study, with an overall attrition rate of 31.8%, the mean reduction in HAM-A total score was −14.7 for patients treated with duloxetine 20mg, −15.3 for patients on duloxetine 60–120mg, and −15.5 for patients in the venlafaxine XR group. The response and remission rates were also similar for the different treatment groups (60 percent vs. 65 percent vs. 61 percent, respectively). Treatment groups did not differ significantly in their rate of study discontinuation due to adverse events.
A poor quality study not included in this report showed results that were consistent with the findings described above. 161
4. SSRIs compared to placebo in adult outpatients with GAD
Sertraline compared with placebo
Currently, sertraline is not FDA-approved for the treatment of GAD. We identified two placebo-controlled trials that assessed the efficacy and tolerability of sertraline in GAD.162–164 Overall these studies found that sertraline could result in better efficacy than placebo in the treatment of GAD.
A 12-week, multicenter, multinational trial randomized 378 outpatients with a primary diagnosis of DSM-IV- defined anxiety disorder to sertraline 50–150 mg/d or placebo. Patients with a history of other psychiatric disorders, including MAD, were excluded. The primary efficacy measure was the HAM-A; secondary assessments included the CGI-I, CGI-S, MADRS, HADS, Q-LES-Q, the Endicott Work Productivity Scale, and the HAM-A psychic and somatic anxiety factors. At endpoint, the mean reduction in HAM-A total score was −11.7 for the sertraline group and −8.0 for the placebo (P<0.0001). Additionally, sertraline was significantly better than placebo on all secondary assessments, including the quality-of-life and work productivity measures.
A 10-week, multicenter, multinational trial randomized 326 outpatients with a primary diagnosis of DSM-IV- defined anxiety disorder to sertraline 50–2000 mg/d or placebo. The inclusion/exclusion criteria were similar to those above as were the outcomes. At endpoint, the mean reduction in HAM-A total score was −12.71 for the sertraline group and −11.15 for the placebo (P=0.032). Additionally, sertraline was significantly better than placebo on secondary assessments, including the quality-of-life and CGI measures.
5. Summary of the evidence
Overall, the evidence is too limited to draw firm conclusions about the comparative efficacy of one second-generation antidepressant to another for treating GAD.
We did not identify any study with a high degree of generalizability.
One RCT reported a greater reduction of HAM-A for escitalopram than paroxetine. This finding, however, is limited to one study funded by the makers of escitalopram.158
The evidence for the comparison of paroxetine with venlafaxine is limited to one poor study and, therefore, insufficient to draw conclusions. 159
B. Obsessive-Compulsive Disorder
The FDA has approved the following SSRIs for the treatment of OCD: fluoxetine, fluvoxamine, paroxetine, and sertraline.
Three head-to-head trials addressing the use of SSRIs or other second-generation antidepressants met our inclusion criteria for the review of OCD (Table 13) One of these head-to-head trials had a 12-week extension phase in which nonresponders were switched to the alternative treatment.167 One additional trial compared citalopram plus mirtazapine to citalopram alone.168 Additionally, one placebo-controlled trial was included because it evaluated an SSRI not covered in the reviews or approved by the FDA (Table 15). Four meta-analyses pooled data from studies comparing SSRIs to placebo. All systematic reviews included comparisons of fluoxetine, fluvoxamine, and sertraline to placebo.169–171 In addition, two reviews included a comparison of paroxetine to placebo170 and one included placebo compared with citalopram.172
Generally, inclusion was based on a criteria-based diagnosis (DSM-III, DSM-IV) of OCD and a predefined cut-off point on an accepted obsessive-compulsive scale (e.g., Y-BOCS, NIMH-OC). The majority of patients could be labeled as having moderate or severe disease with mild or no comorbid depression. Multiple studies limited inclusion by duration of current illness of 1 year or more.
Commonly examined outcome measures were response rate (e.g., more than 25% or 35% improvement of symptoms on an obsessive-compulsive rating scale, or much or very much improved as assessed by a global assessment method), rate of remission (e.g., reduction below a predefined cut-off point on an obsessive-compulsive scale), or changes in score on obsessive-compulsive scales. Comorbid depression or anxiety and quality of life occasionally were assessed as secondary outcome measures.
All included trials could be characterized as efficacy studies. In addition to efficacy, one head-to-head trial specifically evaluated quality of life. Drug or dosing equivalency was present across all trials.
1. SSRIs compared to SSRIs in adult outpatients with OCD
Sertraline compared with fluoxetine
A multicenter Canadian study evaluated the use of sertraline (50–200 mg/d) and fluoxetine (20–80 mg/d) in 150 patients over a 24-week period.173 More than 79 percent of patients had a duration of illness of 10 years or more. Loss to follow-up was 29 percent, with no differential between fluoxetine- and sertraline-treated groups. At 24 weeks, mean response (Y-BOCS) did not differ significantly between the groups, although sertraline-treated patients had shown statistically greater improvement in mean change from baseline (Y-BOCS) at weeks 4, 8, and 12. Remission rates were greater for sertraline-treated patients at week 12 but not at week 24. Both sertraline and fluoxetine showed equivalent efficacy in improving secondary symptoms of depression (HAM-D) and generalized anxiety (CAS). No significant differences in the incidence of side effects between groups were reported.
2. Other second-generation antidepressants compared to SSRIs in adult outpatients with OCD
Venlafaxine compared with paroxetine
A 12-week Dutch study evaluated the use of venlafaxine XR (75–300 mg/d) and paroxetine (15–60 mg/d) in 150 patients.174 Loss to follow-up was 33 percent. At 12 weeks, efficacy as reported by the mean reduction in Y-BOCS total score did not differ significantly between the two groups. Analysis of Y-BOCS obsessions and compulsions subscales revealed an equally high treatment effect over time. Also, response rates (full response ≥ 50% reduction in Y-BOCS; partial response ≥ 35% reduction in Y-BOCS) did not differ at the end of the trial. Quality of life was assessed using the Lancashire Quality of Life Profile: extended Dutch version (LqoLP). Both groups improved on all domains following treatment without showing a significant difference. Incidence rates of insomnia and dry mouth in venlafaxine-treated patients were more than double those in paroxetine-treated patients.
In one head-to-head trial, after a 4-week tapering phase the investigators switched 43 nonresponders to 12 weeks of therapy with the alternate treatment.167 At the end of 12 weeks, intention-to-treat analysis demonstrated a mean decrease on the Y-BOCS of 1.8 in the venlafaxine group and 6.5 in the paroxetine group. Responder rates (Y-BOCS) were 56 percent for paroxetine and 19 percent for venlafaxine; 42 percent of the nonresponders benefited from the crossover.
Escitalopram compared with paroxetine
A 24-week multinational study compared escitalopram (10 or 20 mg/day), paroxetine (40 mg/day and placebo in 466 patients.175 Attrition was 29 percent. At 12 (primary outcome) or 24 weeks, efficacy as reported by the mean reduction in Y-BOCS total score did not differ significantly between the two active groups, nor did the response rates (either CGI-I = 1 or 2 or > 25% Y-BOCS decrease) differ between paroxetine or escitalopram groups.
3. SSRIs augmentation compared to SSRI alone in adult outpatients with OCD
A 12-week trial assessed the additional benefits of augmenting treatment with citalopram (40–80 mg/d) with mirtazapine (15–30 mg/d) in 49 outpatients with OCD.168 Patients were randomized to citalopram plus placebo or citalopram plus mirtazapine. Obsessive-compulsive symptoms were measured with the Y-BOCS; secondary outcome measures included the HAM-D and CGI-I. Loss to follow-up was 8 percent. At endpoint, no significant differences were reported between the two treatment groups. Patients augmented with mirtazapine had a significantly greater reduction in Y-BOCS total score beginning at week 2, although this difference persisted only through week 6 of the study.
4. SSRIs compared to placebo in adult outpatients with OCD
Four meta-analyses reviewed available evidence from placebo-controlled studies;169–172 we rated these analyses as fair quality and one as good quality.172 One study pooled results from 10 trials that compared SSRIs as a class with placebo.169 Data representing 1,076 patients were pooled to define the SSRI group, which consisted of fluvoxamine (five studies), fluoxetine (two studies), and sertraline (three studies). Several studies incorporated multiple dosing arms in the study design.176, 177 For these trials, only the highest dosing arm was incorporated in the meta-analytic results.
As a class, SSRIs were found to be superior to placebo. For obsessive-compulsive symptoms considered together, an effect size of 0.47 (95% CI 0.33 to 0.61) was observed for SSRIs compared to placebo. Considering obsessions and compulsions rated separately, effect sizes were reported as 0.54 (95% CI 0.34 to 0.74) and 0.52 (95% CI 0.34 to 0.70), respectively. Effect sizes generally were consistent for each of the SSRIs when compared to placebo.
A second meta-analysis evaluated placebo-controlled trials of fluvoxamine, fluoxetine, sertraline, and paroxetine.170 Specifically, this study used meta-regression to identify sources of heterogeneity in these trials (and clomipramine trials). They identified 12 trials published before 2000 that compared SSRIs to placebo. Only studies that assessed efficacy with Y-BOCS were incorporated in the meta-regression. Effect sizes were estimated as the difference in improvement (decrease in Y-BOCS) between active drug and placebo.
Four fluvoxamine studies178–181 showed a net improvement of −4.84 (95% CI −7.78 to −1.83). For the three fluoxetine studies,182–184 net improvement was −1.61 (95% CI −2.18 to −1.04); for four sertraline studies,185–188 the pooled difference in Y-BOCS was calculated to be −2.47 (95% CI −6.13 to 1.20). Only one paroxetine study was included; the difference in improvement was estimated as −3.00 (95% CI −4.91 to −1.09).
A third meta-analysis assessed medication effect sizes in six published placebo-controlled trials;171 two fluvoxamine studies;178, 179 two sertraline studies;185, 186 and two fluoxetine studies.182, 183 Compared to placebo, effect sizes did not differ significantly between the three SSRIs evaluated.
A fourth meta-analysis included 17 studies and 3097 participants.172 All consisted of placebo comparisons: five used sertraline, five fluvoxamine, three compared fluoxetine, three paroxetine and one used citalopram. Overall, the drugs evaluated provided greater efficacy than placebo, however, there were differences in the incidence of adverse events, in particular nausea. Three - citalopram, fluvoxamine and paroxetine - had a greater rate of nausea compared to placebo; two - fluoxetine and sertraline - did not.
Citalopram compared with placebo
A fair multicenter study conducted in Europe and South Africa compared various fixed-doses of citalopram to placebo in 401 outpatients with OCD characterized as stable for more than 6 months.177 Loss to follow-up was 16 percent, with small differences between groups. All three doses of citalopram produced significantly more responders (≥ 25% improvement in Y-BOCS) than placebo (P<0.01). The high-dose citalopram (60 mg) response reached statistical significance at week 3, whereas the lower doses (20 mg and 40 mg) reached statistical significance at week 7. On the patient-rated Sheehan Disability Scale, the citalopram-treated patients showed significant improvements for most items. Adverse events were reported in 71 percent of subjects in the active treatment groups. The number of adverse events reported by persons on different citalopram doses did not differ significantly. Ejaculation failure was significantly different from placebo only in the 40 mg citalopram group.
5. Summary of the evidence
Three fair head-to-head studies provide evidence that there is no difference in efficacy between fluoxetine and sertraline or venlafaxine and paroxetine or escitalopram and paroxetine. Other evidence is insufficient to draw conclusions about comparative efficacy between one second-generation antidepressant and another.
We did not identify any study with a high degree of generalizability.
Three head-to-head trials173, 174, #2557 and four meta-analyses169, 170: #3187, 171 provide fair evidence that no difference in efficacy among evaluated second-generation antidepressants exists. Two head-to-head trials provide fair evidence that the efficacy of venlafaxine XR and paroxetine does not differ in improving health outcomes;174, 175, 189 in a follow-up study, 42 percent of nonresponders who switched to the alternative treatment achieved a response.167 One fair placebo-controlled study showed a significantly greater improvement in disability for citalopram compared to placebo.177 In a second study, citalopram-treated patients augmented with mirtazapine had a faster response than patients treated with citalopram alone, although differences did not persist past 6 weeks.168
One study provides fair evidence that sertraline has a faster onset of action than fluoxetine173 in the treatment of OCD. Another fair-rated study reported a faster response for venlafaxine XR compared to paroxetine.174 A fair-rated study showed no difference between escitalopram and paroxetine throughout 24 weeks of treatment.175
FDA-approved evidence exists for the general efficacy of fluoxetine, sertraline, paroxetine, and fluvoxamine for treating OCD. Evidence is insufficient about the efficacy of mirtazapine, bupropion, and nefazodone for treating OCD. Additionally, one study provides fair evidence supporting a greater efficacy of citalopram than placebo.177
C. Panic Disorder
Only fluoxetine, paroxetine, sertraline, and venlafaxine are currently approved by the FDA for the treatment of panic disorder. We viewed FDA approval as evidence for general efficacy and did not review placebo-controlled trials of fluoxetine, paroxetine, sertraline, and venlafaxine.
For panic disorder, we identified four head-to-head trials of fair quality comparing one SSRI, or other second-generation antidepressant to another.190–194 We excluded one study – a single-blinded RCT with a poor quality rating for internal validity191– from our findings, but we discuss it here briefly because of the minimal amount of published research on this topic. Furthermore, we identified three placebo-controlled trials assessing the efficacy and tolerability of fluvoxamine.195–197
Inclusion was generally determined by a criteria-based (DSM-III-R, DSM-IV) diagnosis of panic disorder in addition to a predefined frequency of weekly panic attacks. Patients with at least one to four panic attacks per week or eight in total over the past 4 weeks were eligible for inclusion. Both patients with and without agoraphobia were included in these trials. Common exclusion criteria were additional Axis I disorders, high suicidal risk, a history of alcohol or drug dependence or abuse, use of other psychotropic medications, and progressive medical disease.
The primary outcome measure in all trials was the frequency of panic attacks as assessed with various scales (e.g., Panic and Agoraphobia Scale, Modified Panic and Anticipatory Anxiety Scale [PAAS], Panic Associated Symptoms Scale [PASS]). Secondary outcome measures included changes from baseline in the Panic Disorder Severity Score (PDSS), quality of life and health-related functional capacity (Sheehan Disability Scale [SDS], Fear Questionnaire [FQ]), the Phobia Scale, anxiety-related subscales of the MADRS and HAM-D, and global assessment methods (e.g., CGI, Q-LES-Q).
1. SSRIs compared to SSRIs in adult outpatients with Panic Disorder
Four fair double-blinded RCTs compared the efficacy and tolerability of one SSRI to another.
Citalopram compared with escitalopram
One multicenter study randomized 366 patients with panic disorder to citalopram (10–40 mg/d), escitalopram (5–20 mg/d), or placebo.190 Study duration was 10 weeks. Patients with and without concomitant agoraphobia were included. Quality of life and health-related functional capacity were additional outcome measures. Loss to follow-up was 32 percent. The frequency of panic attacks was significantly reduced for escitalopram compared to placebo (P=0.04) but not for citalopram compared to placebo. Both treatments significantly improved quality of life, panic disorder symptoms, and severity of the disease (P<0.05) compared to placebo. The article does not report a direct comparison of citalopram to escitalopram; presumably the two active treatment groups did not differ significantly on efficacy measures.
Sertraline compared with paroxetine
A German RCT randomized 225 patients with panic disorder to paroxetine (40–60 mg/d) or sertraline (50–150 mg/d).192 Study duration was 12 weeks. Patients with and without concomitant agoraphobia were included. Quality of life was assessed as a secondary outcome measure. Results revealed no statistically significant differences in PAS (Panic and Agoraphobia Scale) scores between treatment groups (P=0.589). Furthermore, no statistical differences in secondary outcome measures (PAS subscales, CGI-S, HAM-A, Sertraline-Quality of Life Battery) could be detected.
Citalopram compared with paroxetine
A small Italian trial enrolled 58 patients to citalopram (20–50 mg/d) and paroxetine (20–50 mg/d) for 60 days.191 Patients and care providers were not blinded to treatment allocation; therefore, this study received a poor quality rating for internal validity. Loss to follow-up was 10 percent. Results reported no statistically significant differences between citalopram and paroxetine in any efficacy measures. However, results may be biased because of lack of double blinding.
Venlafaxine ER compared with paroxetine
Two multi-national fixed-dose RCTs compared two different doses of venlafaxine ER to paroxetine (venlafaxine ER 75 mg/d or 150 mg/d compared with paroxetine 40 mg/d and venlafaxine ER 75 mg/d or 225 mg/d compared with paroxetine 40 mg/d).193, 194 Both studies received a fair rating for internal validity. Loss to follow up was reported as 20.8 percent and 20.1 percent, respectively. Results provided mixed findings. The study conducted in Europe (N=664) demonstrated no statistically significant difference in efficacy between venlafaxine ER 75 mg/d or 150 mg/d and paroxetine 40 mg/d (patients free from full-symptom panic attacks at 12 weeks: 54.4% compared with 59.7% compared with 60.9%).193 In the second trial (N=653), the venlafaxine ER 225 mg/d group had a significantly greater percentage of patients free of full-symptom panic attacks at the 12 week endpoint compared to the paroxetine 40 mg/d group (70.0% compared with 58.3%; P<0.05) and also had a significantly lower PDSS score (4.78 compared with 6.26; P<0.05).194 However, this study compared a high dose of venlafaxine ER to a medium dose of paroxetine.
2. SSRIs compared to placebo in adult outpatients with Panic Disorder
Fluvoxamine compared with placebo
Three fair-rated studies, all lasting 8 weeks, compared fluvoxamine (50–300 mg/d) to placebo.195–197 The first study enrolled 75 patients to fluvoxamine (50–300 mg/d), placebo, or cognitive therapy.195 Loss to follow-up was 20 percent. Outcome measures included functional capacity (Sheehan Disability Scale). Statistical analysis did not fulfill accepted criteria for intention-to-treat analysis (only subjects who completed 3 weeks of medication were analyzed). Fluvoxamine showed significantly greater improvements in all primary (Panic Attack Severity Score, Clinical Anxiety Score [CAS], CGI, MADRS) and secondary (Sheehan Disability Scale) efficacy measures compared to placebo.
The second study randomized 50 patients to fluvoxamine (50–300 mg/d) or placebo.196 Loss to follow-up was 28 percent, and no intention-to-treat analysis was done. The fluvoxamine group reported significantly fewer major panic attacks starting at week 4 until the endpoint (P<0.05); they also had significantly lower scores on CAS and MADRS (P<0.05). By contrast, active drug and placebo groups did not differ significantly in terms of minor panic attacks and Sheehan disability scores.
The third trial enrolled 188 participants.197 Loss to follow-up was about 35 percent. Results were consistent with the other studies. Fluvoxamine showed a significantly greater efficacy in most primary (Daily Panic Attack Inventory) and secondary (MADRS, CGI-I, CGI-S, CAS, Sheehan Disability Scale) outcome measures compared to placebo.
3. Summary of the evidence
Two fair fixed-dose trials provide inconclusive evidence on the comparative efficacy of venlafaxine ER and paroxetine. One fair head-to-head study provides evidence that efficacy does not differ between citalopram and escitalopram. In other trials, significant differences in study design and outcome selection make this evidence insufficient to identify differences between treatments.
We did not identify any study with a high degree of generalizability.
While one fair RCT showed venlafaxine ER 225 mg/d to be superior to paroxetine 40 mg/d in reducing full-symptom panic attacks and in PDSS score,194 the same effect was not seen when comparing venlafaxine ER 150 mg/d or 75 mg/d and paroxetine 40 mg/d.193, 194 Two fair RCTs provide evidence that the efficacy of reducing panic attacks and improving quality of life does not differ significantly between citalopram and escitalopram190 or between paroxetine and sertraline192 in outpatients with panic disorder. Fair evidence exists from three placebo-controlled trials of significantly greater efficacy and improvement of health outcomes and functional capacity for fluvoxamine compared to placebo.196–199 FDA-approved evidence supports the general efficacy of fluoxetine, paroxetine, venlafaxine and sertraline for the treatment of panic disorder. Evidence is insufficient about the efficacy of duloxetine, mirtazapine, bupropion, and nefazodone for treating panic disorder.
D. Post-Traumatic Stress Disorder
Currently, only paroxetine and sertraline have been FDA-approved for the treatment of post-traumatic stress disorder (PTSD). As in other chapters, we view FDA-approval as evidence for general efficacy and, therefore, do not review placebo-controlled trials on drugs that have been FDA-approved.
For PTSD, we found four head-to-head studies: one comparing citalopram to sertraline,200 two comparing nefazodone to sertraline,201, 202 and one comparing venlafaxine to sertraline.203 No other second-generation antidepressants were compared to one another.
In addition we included four placebo-controlled trials assessing the efficacy of fluoxetine and venlafaxine, which are not FDA-approved for the treatment of PTSD (Table 17).
Inclusion of patients was generally determined by a criteria-based (DSM-III-R, DSM-IV) diagnosis of PTSD in addition to a predefined threshold on a universally used PTSD scale (Clinician Administered PTSD Scale [CAPS]). The majority of patients had suffered physical or sexual abuse or had witnessed injury or death of a third person. More than half of the participants had a concomitant diagnosis of MDD or GAD or a history of alcohol and substance abuse.
1. SSRIs compared to other second-generation antidepressants in adult outpatients with PTSD
Sertraline compared with Citalopram
A fair study randomized 59 outpatients with PTSD to 10 weeks of citalopram (20–50 mg/d), sertraline (50–200 mg/d), or placebo.200 Primary outcomes measures (CAPS, BDI) did not indicate any statistically significant differences in efficacy between citalopram and sertraline and between the active treatments and placebo.
Sertraline compared with Nefazodone
A fair-rated RCT randomized 37 patients with PTSD to 12 weeks of sertraline (50–200 mg/d) or nefazodone (100–600 mg/d).201 Sertraline- and nefazodone-treated patients did not differ significantly on primary (CAPS2, CGI) and secondary outcome measures (DTS, MADRS, PSQI, SDS, HAM-A). Both treatment groups had statistically significant improvements within group from baseline to endpoint on all outcome measures. Loss to follow-up was 38 percent; the rate of post-randomization exclusion because of lack of data was 28 percent. Results of this study were consistent with findings from an open-label trial in Turkish earthquake survivors.202 This study met our formal eligibility criteria; however we determined it to be of poor quality (completers analysis only). Because of the lack of head-to-head evidence we are including its findings. Sixty earthquake survivors received sertraline or nefazodone in a non-randomized manner, based on availability. No differences in efficacy outcomes (Posttraumatic Stress Diagnostic Scale [PDS], Posttraumatic Stress Disorder Scale [TOP-8], CGI) could be detected between patients on sertraline or nefazodone after 6 months of treatment.
Sertraline compared with Venlafaxine
A fair 12-week, placebo-controlled RCT (N=538) evaluated the comparative efficacy and safety of sertraline (25–200 mg/d) and venlafaxine ER (37.5–300 mg/d).203 At study endpoint, 30.2 percent on venlafaxine ER and 24.3 percent on sertraline achieved remission. In other primary outcome measures the efficacy of sertraline and venlafaxine ER was similar (CAPS, CGI-S, Assessment of Functioning [GAF], Vulnerability to the Effects of Stress Scale [SVS]). Both treatment groups had statistically significant improvements on all outcome measures compared with placebo.
2. SSRIs compared to placebo in adult outpatients with PTSD
Fluoxetine compared with placebo
Three placebo-controlled RCTs provide conflicting results on the general efficacy of fluoxetine for the treatment of PTSD.204, 205 A small fair-rated study enrolled 54 patients to 12 weeks of fluoxetine (10–60 mg) or placebo.204 Loss to follow-up was 31.5 percent. Using the Duke Global Rating for PTSD cut-off score of 1 (no symptoms) to define responders, the fluoxetine group had significantly more responders than the placebo group (59% compared with 19%; P<0.005). According to Duke Global Rating for PTSD cut-off scores of 1 (no symptoms) or 2 (minimal symptoms) to define responders, a nonstatistically significant trend toward fluoxetine was observed (P=0.06). Health-related secondary outcome measures (SIP, disability and stress subscales) showed significantly greater improvements for fluoxetine (P<0.005). A Kaplan-Meier analysis reported a significantly faster onset of efficacy for fluoxetine (P<0.005) than for placebo.
Two additional, fair studies did not detect any statistically significant differences between fluoxetine and placebo for the treatment of PTSD. One study was a 12-week, fixed-dose (fluoxetine 20 or 40 mg/d) trial (N=411) that enrolled primarily women (71%) with PTSD.205 At study endpoint both primary outcome measures (TOP-8, CAPS) showed similar efficacy outcomes between fluoxetine and placebo. The other trial (N=88) was an 8-week flexible-dose RCT that compared fluoxetine (20–60 mg/d) to placebo, psychotherapy, or eye movement desensitization and reprocessing.206 No significant differences in CAPS scores were detected at endpoint between fluoxetine- and placebo-treated patients.
Venlafaxine compared with placebo
A fair, 6-month, placebo-controlled RCT assessed the efficacy of venlafaxine ER (37.5–300 mg/d) in 329 patients with PTSD.207 Overall improvements were significantly greater for patients on venlafaxine ER than on placebo (CAPS, CGI-S, HAM-D). After 6 months, 51 percent of patients on venlafaxine ER achieved remission compared with 38 percent on placebo (P=0.01). Patients on venlafaxine ER had also greater improvements than the placebo group with respect to quality of life and functional capacity. Withdrawal rates were similar between groups.
3. Summary of the evidence
We identified one head-to-head trial comparing citalopram to sertraline, one study comparing sertraline to nefazodone and one study comparing sertraline to venlafaxine.
We did not identify any study with a high degree of generalizability.
Three head-to-head trials did not detect any differences in efficacy between citalopram and sertraline,200 sertraline and nefazodone,201 and sertraline and venlafaxine ER.203 FDA-approved evidence exists for the general efficacy of paroxetine and sertraline for treating PTSD. Placebo-controlled trials report general efficacy of venlafaxine but not of fluoxetine in the treatment of PTSD. Significant differences in population characteristics make this evidence insufficient to identify differences between treatments based on placebo-controlled evidence.
E. Social Anxiety Disorder
Currently, three SSRIs – fluvoxamine CR, paroxetine and sertraline – are approved by the FDA for the treatment of social anxiety disorder. In addition, the extended release formulation of one SNRI – venlafaxine – is approved for the treatment of social anxiety disorder.
Three head-to-head trials (with placebo arms) compared one second-generation antidepressant to another for the treatment of social anxiety disorder.208–210 Two 12-week trials compared paroxetine to venlafaxine ER;208, 210 a 24-week trial compared escitalopram to paroxetine.209 All three trials included measures of functional capacity in addition to efficacy and tolerability.
We reviewed additional evidence from placebo-controlled trials if they assessed a second-generation antidepressant not currently FDA-approved for social anxiety disorder. One meta-analysis compared fluvoxamine, paroxetine, and sertraline to placebo,211 an additional meta-analysis summarized the comparative evidence and conducted indirect comparisons of second-generation antidepressants using network-analysis,212 and one systematic review compared SSRIs to placebo.213 In addition, 6 placebo-controlled studies evaluated second-generation antidepressants currently not approved by the FDA for social anxiety disorder: two escitalopram studies,214, 215 two fluoxetine studies,216, 217 one mirtazapine study,218 and one nefazodone study.219 (Table 18).
In general, inclusion was based on a criteria-based diagnosis (DSM-IV) of social anxiety disorder. Several studies required a minimal duration of current illness of 6 months or greater.208, 210, 216, 219 Additionally, several studies limited eligibility using a predefined cut-off point on a validated anxiety rating scale.208–210, 215, 216
The main outcome measures examined were mean change in anxiety as measured by one of several scales, including the Liebowitz Social Anxiety Scale (LSAS), the Brief Social Phobia Scale (BSPS), the HAM-A, and the social phobia subscale of the Marks Fear Questionnaire (MF). Social anxiety global assessment scales such as the Clinical Global Impression-Social Phobia Scale (CGI-SP) also were used. Several studies included patient-rated measures of anxiety using the Social Phobia Scale (SPS) or the Social Phobia Inventory (SPI). Disability, health status, quality of life, and comorbid depression were frequently assessed as secondary outcome measures.
Trial reporting was often incomplete. All trials used an intention-to-treat analysis. Among the included studies, loss to follow-up was between 20 percent and 36 percent. One study had a loss-to-follow-up differential between treatment groups greater than 10 percentage points (13.8 points).219
All included trials are characterized as efficacy studies. One study assessed relapse prevention randomizing escitalopram responders (CGI-I score of 1 or 2) to 24 weeks of escitalopram or placebo.214 This study evaluated the rate of relapse between active treatment and placebo.
1. SSRIs compared to SSRIs in adult outpatients with social anxiety disorder
One fair-rated double-blinded RCT compared the efficacy and tolerability of one SSRI to another. In addition, a meta-analysis conducted indirect comparisons of second-generation antidepressants for the treatment of social anxiety disorder.
Escitalopram compared with paroxetine
One multinational study randomized 839 patients with social anxiety disorder to fixed doses of escitalopram (5, 10, or 20 mg/d), paroxetine 20 mg/d, or placebo.209 Eligible patients had a baseline LSAS score of 70 or higher with a score of 5 or higher on one or more of the SDS subscales. Overall loss to follow-up in this 24-week trial was 29 percent. The primary outcome measure was mean change from baseline to week 12 in the LSAS total score; secondary outcome measures included the subscales, Clinical Global Impression of Improvement scale (CGI-I), Clinical Global Impression of Severity scale CGI-S, and the Sheehan Disability Scale (SDS). No significant differences in LSAS total score were observed between any escitalopram treatment group and the paroxetine group in the intention-to-treat analysis. The authors did not report any intention-to-treat results for secondary outcome measures. In the observed-cases-analysis at 24 weeks, escitalopram 20 mg/d was superior to paroxetine 20 mg/d on the CGI-S. Significant differences (favoring escitalopram 20 mg/d) were noted on the SDS at weeks 16 and 20, but differences between escitalopram and paroxetine were not significantly different at week 24.
Indirect comparisons of escitalopram, fluvoxamine, paroxetine, and sertraline
A good meta-analysis of second-generation antidepressants for social anxiety disorder utilized data of more than 6500 patients from three head-to-head trials and 15 placebo-controlled trials. To determine the comparative efficacy among drugs, authors employed network meta-analyses.212 With the exception of one study, which included children and adolescents, trial populations consisted of adults with mean ages from 35 to 41 years and a relatively equal distribution of males and females. Baseline disease severity varied among participants (range of LSAS scores 74–97). Trials included in the analysis had to have a minimum duration of 12 weeks (range of study duration 12–28 weeks). Individual drugs were included in the network meta-analysis when at least two similarly designed trials provided CGI-I data. Authors conducted a network-meta-analysis and found no significant differences in response among included SSRIs.
Because of the limited number of component studies, however, estimates of relative effects were imprecise with wide confidence intervals which encompassed potentially important differences.
2. SNRIs compared to SSRIs in adult outpatients with social anxiety disorder
A good meta-analysis conducted indirect comparisons of second-generation antidepressants for the treatment of social anxiety disorder. Two fair double-blinded RCTs compared the efficacy and tolerability of one second-generation antidepressant to an SSRI. An additional
Indirect comparisons of venlafaxine with SSRIs
The above mentioned good meta-analysis of second-generation antidepressants for social anxiety disorder conducted indirect comparisons of venlafaxine with various SSRIs (escitalopram, fluvoxamine, paroxetine, and sertraline) using network-meta-analysis of data on more than 6500 patients three head-to-head trials and 15 placebo-controlled trials.212 The authors found no significant differences in any of the possible comparisons between venlafaxine and escitalopram, fluvoxamine, paroxetine, or sertraline. However, estimates had wide confidence intervals and encompassed potentially important differences.
Venlafaxine compared with paroxetine
Two 12-week multicenter trials compared venlafaxine ER to paroxetine and placebo.208, 210 A European trial randomized 436 patients with social anxiety disorder208 and an American trial randomized 440 patients with social anxiety disorder210 to venlafaxine ER (75–225 mg/d), paroxetine (20–50 mg/d), or placebo. Eligible patients were 18 years or older who met DSM-IV criteria for social anxiety disorder at least 6 months before enrollment. In the European trial, significantly more females were randomized to placebo than to venlafaxine or paroxetine. The primary outcome measure was the LSAS; secondary outcome measures included the CGI-I, CGI-S, SPI, and SDI. The European trial also included a measure of work productivity WPAI. At 12 weeks, no significant differences in any outcome measure were observed between venlafaxine ER and paroxetine in either trial. Both venlafaxine ER and paroxetine were significantly better than placebo for all primary and secondary outcome measures (P<0.05), including the measures of functional capacity (SDI) and work productivity (WPAI).
3. SSRIs compared to placebo in adult outpatients with social anxiety disorder
One meta-analysis, one systematic review, and five placebo-controlled trials provide additional evidence.
4. SSRIs compared with placebo
One systematic review evaluated the efficacy of SSRIs compared with placebo in the treatment of social anxiety disorder in adults.213 This review included placebo-controlled trials of SSRIs ranging in duration from 10–24 weeks and converted treatment effects to standardized effect sizes. Authors concluded that, in general, SSRIs are more effective than placebo in treating social anxiety disorder.
Escitalopram compared with placebo
One fair 12-week study compared flexible doses of escitalopram to placebo.215 This trial randomized 358 participants meeting DSM-IV criteria for social anxiety disorder with a score of at least 70 on the LSAS to escitalopram (10–20 mg/d) or placebo. Overall loss to follow-up was 19 percent (18% for placebo and 20% for escitalopram). The primary efficacy measure was the LSAS total score; secondary outcome measures included the LSAS subscales, CGI-S, CGI-I, SDS, and MADRS. At endpoint, escitalopram was significantly better than placebo as assessed by the LSAS total score (P<0.01), LSAS subscales (P<0.05), CGI-S (P<0.01), CGI-I (P<0.01), and the work and social domains of the SDS (P<0.05). Results were similar to the placebo comparison reported by Lader et al.209 The most common adverse event reported for escitalopram or placebo was headache (25% in both groups); compared to placebo, more patients randomized to escitalopram reported nausea (12% compared with 22%; P=NR).
One fair relapse prevention study openly treated 517 patients with generalized social anxiety disorder with escitalopram (10–20 mg/d) for 12 weeks.214 Responders (CGI-I score of 1 or 2) were randomized to 24 weeks of double-blind treatment with escitalopram or placebo. The primary efficacy parameter was time to relapse, defined as ≥ 10 point increase in LSAS total score from randomization. Of 372 randomized patients, 198 escitalopram-treated patients (65%) and 75 placebo-treated patients (41%) completed the 24-week study. In the escitalopram group, 42 patients relapsed (22%), while 91 patients (50%) relapsed in the placebo group. The median time to relapse was 407 days for escitalopram-treated patients and 144 days for placebo-treated patients (P<0.001).
Fluoxetine compared with placebo
Two fair studies compared flexible doses of fluoxetine to placebo.216, 217 The first trial randomized 60 participants meeting DSM-IV criteria for social anxiety disorder for at least 6 months to 14 weeks of fluoxetine (20–60 mg/d) or placebo. Loss to follow-up was 20 percent with a higher rate in the placebo control group than the active fluoxetine group (23% compared with 16%, respectively). The primary efficacy measure was the LSAS. Significant improvements in LSAS scores were reported for fluoxetine and placebo, with no statistically significant differences between groups (P=0.901). Secondary efficacy measures included the BSPS, FQ, HAM-A, HAM-D, Global Assessment of Functioning (GAF), and SF-36. Overall, no statistically significant differences were reported on secondary efficacy measures. Compared to placebo, fluoxetine-treated patients had a significant increase in the bodily pain subscale of the SF-36 (P=0.05). Significantly more fluoxetine-treated patients had asthenia than placebo-treated patients (P<0.05).
The second trial217 randomized 117 patients meeting DSM-IV criteria for social anxiety disorder (no minimum time of illness) to fluoxetine (10–60 mg/d) or placebo for 14 weeks. (In total, 295 patients were randomized in this study to arms that included comprehensive cognitive behavioral therapy. However, we included only two arms—the fluoxetine arm and the placebo arm.) The attrition rate was 36 percent with a higher rate in the placebo group than the fluoxetine group (40% compared with 32%); however, the differential rate was not considered high. Primary efficacy measures were the CGI-I, CGI-S and BSPS. CGI-I response rates were significantly higher in fluoxetine treated patients (51% compared with 32%). Fluoxetine-treated patients also showed a significantly greater improvement in CGI-S score from baseline (P<0.05) and in Social Phobia and Anxiety Inventory (SPAI) score (P<0.05).
5. Other second-generation antidepressants compared with placebo
Mirtazapine compared with placebo
One fair 10-week trial compared mirtazapine to placebo in 114 women with social phobia.218 The primary outcome measure was the change in SPIN score; LSAS and SF-36 scores also were assessed. After 10 weeks, mirtazapine-treated patients were significantly more improved than placebo-treated patients on the SPIN (difference in change = −8.1; P<0.001), LSAS (difference in change −20.2; P<0.001), and the SF-36 domains of general health perception, vitality, social functioning, role-emotional, and mental health (P<0.001 for all). Statistically significant differences were not noted in physical functioning (P=0.91), role-physical (P=0.77), and bodily pain (P=0.53).
Nefazodone compared with placebo
One fair trial compared nefazodone to placebo in adults meeting the DSM-IV criteria for general social phobia for at least 1 year.219 105 patients were randomized to nefazodone (100–600 mg/d) or placebo for 14 weeks. The primary outcome measures were percentage of CGI-I responders (1 or 2) at endpoint and the mean change from baseline in LSAS total score. Secondary efficacy measures included CGI-S, Social Phobia Inventory, SPS, and Social Interaction Anxiety Scale. More nefazodone- than placebo-treated patients were CGI-I responders, but the difference was not significant (31.4% compared with 23.5%, P=0.38). With the exception of the Social Phobia scale, there were no significant differences between groups in measures of social phobia. Nefazodone-treated patients had significantly higher incidences of some adverse events: dizziness (P<0.01), nausea/vomiting (23.5% compared with 7.8%, P=0.03), and dry mouth (23.5% compared with 2.0%, P<0.01).
6. Summary of the evidence
Three head-to-head trials compared one second-generation antidepressant to another for the treatment of social anxiety disorder. These trials suggest no differences in efficacy for escitalopram compared with paroxetine and venlafaxine ER compared with paroxetine. These findings were confirmed in a network meta-analysis that did not find any significant differences in any of the possible comparisons between venlafaxine ER, escitalopram, fluvoxamine, paroxetine, or sertraline. Additionally, indirect evidence from a meta-analysis of placebo-controlled trials provides evidence that there is no difference in efficacy between fluvoxamine, paroxetine, and sertraline.
We did not identify any study with a high degree of generalizability.
One comparative trial provides fair evidence of comparable efficacy between escitalopram and paroxetine for the treatment of social anxiety disorder.209 Two comparative trials provide fair evidence of comparable efficacy between venlafaxine ER and paroxetine.208, 210 One meta-analysis of placebo-controlled studies provides fair evidence of comparable efficacies of fluvoxamine, paroxetine, and sertraline for the treatment of social anxiety disorder.211
One network meta-analysis of head-to-head trials and placebo-controlled studies provides fair evidence of comparable efficacy between escitalopram, fluvoxamine, paroxetine, sertraline and venlafaxine ER.212 Six trials and one systematic review.213 provide fair evidence that SSRIs significantly improve health outcomes compared to placebo.208–210, 215, 217, 218
Two placebo-controlled trials did not support the efficacy of fluoxetine216 and nefazodone.219 Evidence from three placebo-controlled trials supports the efficacy of escitalopram,209, 214, 215 and evidence from one placebo-controlled trial supports the efficacy of mirtazapine in women.218 Evidence is insufficient about the efficacy of citalopram, duloxetine, mirtazapine, bupropion, and nefazodone for treating social anxiety disorder.
III. For adult outpatients with premenstrual dysphoric disorder or late luteal phase dysphoric disorder, do SSRIs or second-generation antidepressants differ in efficacy?
The FDA has approved fluoxetine, sertraline, and paroxetine CR for the treatment of premenstrual dysphoric disorder (PMDD) and late luteal phase dysphoric disorder (LLPDD). We did not find any head-to-head studies comparing SSRIs or other second-generation antidepressants to each other. Two systematic reviews 220, 221 and two RCTs222, 223 compared second-generation antidepressants to placebo. These studies are listed in Table 17.
Studies were conducted over two to six menstrual cycles. Some studies included in the meta-analyses 220, 221 compared intermittent luteal phase therapy with continuous treatment and with placebo. Included studies were conducted in women of reproductive age (18 to 49 years) with a clinical diagnosis of PMDD or LLPDD 220 or in women of any age who met the diagnostic criteria for PMS, PMDD and LLDD 221. Women were required to meet DSM criteria in all two trials. The more recent meta-analysis included studies which used Self-Rating scales, confirmation by psychiatric evaluation or predefined diagnostic criteria for PMDD or LLPDD according to DSM-III or DSM-IV.220 The detailed interviews required to determine a diagnosis of PMDD in these studies may limit the generalizability of the findings to patients in others settings such a primary care or gynecological offices where a diagnosis of PMDD is often made on less strict criteria. Most studies excluded women with depression or other psychiatric illness, those with irregular menstrual cycles, and those taking hormones (including oral contraceptives). Both placebo-controlled trials used a patient-assessed daily symptom rating or report in addition to the CGI.222, 223 Patients monitored their symptoms through the use of diaries, calendars, or visual analog scales. In addition to patient reports of symptoms, one trial used the 21-item HAM-D.222 Studies included in the meta-analyses used similar efficacy outcome measures.
1. SSRIs compared to placebo in adult outpatients with premenstrual dysphoric disorder or late luteal phase dysphoric disorder
SSRIs compared with placebo
The updated Cochrane Collaboration Report 220 reported on efficacy outcomes of FDA-approved and non-FDA-approved SSRIs. This good-quality meta-analysis pooled data from 22 trials comparing various SSRIs to placebo, including citalopram, escitalopram, fluoxetine, fluvoxamine, paroxetine, and sertraline.
Citalopram was more effective than placebo with a SMD of −1.27 (95% CI −1.86 to −0.69) P<0.0001. (The three included studies were different arms of one study comparing placebo to citalopram in different dosages.) There was only one study with fluvoxamine and therefore no meta-analysis was conducted. This RCT did not fulfill our inclusion criteria due to the small sample size.
The second systematic review 221 provides consistent results. Citalopram was more effective than placebo (OR: 0.18; 95% CI 0.06 to 0.51).
2. Other second-generation antidepressants compared to placebo in adult outpatients with premenstrual dysphoric disorder or late luteal phase dysphoric disorder
Venlafaxine compared with placebo
One fair RCT compared an SNRI, specifically a continuous daily dose of venlafaxine (50–200 mg/d), to placebo over four menstrual cycles.222 It reported 36 percent of subjects as lost to follow-up. Venlafaxine-treated subjects had significantly lower premenstrual daily symptom report scores and 21-item HAM-D scores than placebo subjects. Sixty percent of venlafaxine-treated subjects were considered responders (e.g., had more than a 50% reduction in baseline symptom report score), whereas only 35 percent of placebo-treated subjects were characterized as responders.
Nefazodone compared with placebo
One fair RCT compared a second-generation antidepressant, specifically both a continuous and intermittent daily dose of nefazodone (100–400 mg/d), to placebo over two menstrual cycles.223 This trial did not, however, compare intermittent and continuous therapy to each other. Twenty-two percent of subjects were reported as lost to follow-up in this trial. For both dosing methods, no significant differences were seen between nefazodone and placebo in either patient self-rated global improvement or any of the individual symptoms assessed (irritability, depressed mood, affect lability, tension, breast tenderness, bloating, and food craving).
3. Summary of the evidence
We identified no head-to-head trials. Significant differences in study characteristics make this evidence insufficient to identify differences among treatments.
We did not identify any study with a high degree of generalizability.
Two meta-analyses provided good evidence that citalopram has a significantly greater efficacy than placebo in the treatment of PMDD and LLPDD. 220, 221 One fair RCT provides evidence that the efficacy is significantly greater for venlafaxine than for placebo.222 Lastly, evidence from one fair RCT indicates that nefazodone does not have greater efficacy than placebo in the treatment of PMDD or LLPDD.223 There is FDA-approved evidence of the efficacy of fluoxetine, paroxetine, and sertraline in the treatment of PMDD and LLPDD. We could not identify sufficient evidence on the efficacy of escitalopram, mirtazapine, and bupropion for treating either PMDD or LLPDD.
Key Question 2. For outpatients with depressive, anxiety, and/or premenstrual dysphoric disorder, do second-generation antidepressants differ in safety, tolerability, or adverse events?
Most of the studies that examined the efficacy of one drug relative to another also determined differences in tolerability. Methods of adverse events assessment differed greatly. Few studies used objective scales such as the UKU-SES (Utvalg for Kliniske Undersogelser Side Effect Scale) or the adverse reaction terminology from the World Health Organization (WHO). Most studies combined patient- reported adverse events with a regular clinical examination by an investigator. Often it was hard to determine whether assessment methods were unbiased and adequate. Rarely were adverse events prespecified and defined. Short study durations and small sample sizes additionally limited the validity of adverse events assessment in many trials.
Few RCTs were designed to assess adverse events as primary outcomes. Most published studies were post hoc analyses or retrospective reviews of databases. We included observational studies if the sample size was larger than 100 and the study duration was at least 1 year (Table 21).
A. Tolerability and Discontinuation Rates
Nausea, headache, diarrhea, fatigue, dizziness, sweating, sexual side effects, tremor, dry mouth, and weight gain were commonly reported adverse events. Overall, second-generation antidepressants led to similar adverse events. The frequencies of specific adverse events, however, differed among some second-generation antidepressants.29, 30, 32, 225
Table 20 depicts the mean incidence and 95% CI for specific adverse events commonly reported in head-to-head trials. Statistics are descriptive only and comparisons across different drugs should be made with caution given differences in assessment and reporting of adverse events across trials.
Venlafaxine had a consistently higher rate of nausea and vomiting than SSRIs. In six studies, the difference reached statistical significance.93, 94, 97, 101, 102, 104 In six additional trials, the higher rates of nausea or vomiting for venlafaxine were not statistically significant.95, 96, 98, 100, 105, 107 The rate of patients reporting nausea or vomiting ranged from 8 percent to 48 percent. A meta-analysis compared the pooled relative risk of nausea and vomiting for venlafaxine with that for comparator SSRIs as a class.225 The RR was 1.53 (95% CI, 1.26–1.86). The corresponding number needed to harm (NNH) was 9 (95% CI, 6–23). In a subgroup analysis authors limited studies to those with extended-release formulations. Pooled results still detected a higher risk of nausea and vomiting for venlafaxine extended-release than for SSRIs but the statistical significance was lost (RR 1.38; 95% CI 0.93–2.05)
A pooled analysis of published and unpublished trials of duloxetine did not find significant differences in nausea between duloxetine (40–120 mg/d) and paroxetine (20 mg/d) or between duloxetine (120 mg/d) and fluoxetine (20 mg/d).226 A meta-analysis of published and unpublished studies of duloxetine compared with escitalopram, fluoxetine, paroxetine, or venlafaxine as a class yielded similar risks for experiencing adverse events (RR 1.22; 95% CI 0.62–2.43).28 Duloxetine, however, led to a significantly higher risk of overall discontinuation (RR 1.57; 95% CI 1.27–1.93) or discontinuation due to adverse events (RR 1.16; 95% CI 1.04–1.30) than the comparator drugs as a class.
In most studies, sertraline led to higher rates of diarrhea than did comparator drugs (bupropion, citalopram, fluoxetine, fluvoxamine, mirtazapine, nefazodone, paroxetine, and venlafaxine).53, 54, 73, 75, 77, 79, 82, 83, 91, 107, 113, 121 Incidence was 8 percent (95% CI, 3–11 percent) higher than with comparator drugs. The NNH was 13 (95% CI, 9–29).225 These results have been confirmed by a Cochrane review. The pooled risk of diarrhea was significantly greater for patients on sertraline than patients treated with bupropion (OR 3.88; 95% CI 1.50–10.07) or mirtazapine (OR 2.74; 95% CI 1.52–4.97).32
Whether this finding can be extrapolated to comparisons of sertraline with other second-generation antidepressants remains unclear.
A British study pooled data from Prescription-Event-Monitoring (PEM) of general practitioners 6 months to 1 year after they had issued prescriptions.227, 228 Included drugs were fluoxetine, fluvoxamine, paroxetine, sertraline, venlafaxine, and nefazodone. The final cohort exceeded 10,000 patients for each drug. Demographics and indications were comparable among study groups. Nausea and vomiting were the two most frequent clinical reasons for withdrawal in the first month of treatment for all drugs. Venlafaxine had the highest rate of nausea and vomiting per 1000 patient months. Like patients using paroxetine, venlafaxine patients also most frequently reported male sexual dysfunction. However, sweating, impotence, and ejaculation failure were significantly higher in the paroxetine group than in the other groups (P=0.004; P<0.001). In addition, patients using paroxetine and those using nefazodone most frequently reported drowsiness and sedation. Sertraline and fluoxetine had significantly lower rate ratios of agitation and anxiety. However, there were more reports of mania during 90 days with fluoxetine than with any other drug. The death and suicide rates did not differ significantly among study groups. Among SSRIs only, drowsiness and sedation were significantly higher in the fluvoxamine and paroxetine group than in the fluoxetine and sertraline group. Overall, the mean incidence density per 1000 patient months for SSRIs was highest for fluvoxamine (fluvoxamine 17.6; fluoxetine 7.0; paroxetine 7.6; sertraline 6.2). Suicide rates did not differ significantly among study groups. Adverse events were reported by physicians rather than patients; the nonresponse rate was 40 percent. Therefore, measurement bias, selection bias, and potential confounding may compromise these results.
Three RCTs were powered primarily to detect differences in adverse events between fluvoxamine and citalopram229 and fluvoxamine and paroxetine,80 and fluvoxamine and fluoxetine.67 A Dutch multicenter trial was designed to assess between-group comparisons of gastrointestinal side effects between citalopram (20–40 mg/d) and fluvoxamine (100–200 mg/d).229 A total of 217 patients were enrolled for 6 weeks. Overall, 57 percent of patients reported adverse events. Significantly more patients in the fluvoxamine group had an excess incidence of diarrhea (+13%; P=0.026) or nausea (+16%; P=0.017). However, the authors did not provide a baseline comparison of gastrointestinal illnesses between groups. Differences at baseline could bias results.
The second study enrolled 60 patients to fluvoxamine (50–150 mg/d) or paroxetine (20–50 mg/d) for 7 weeks.80 Sweating was the only significantly higher adverse event: 30 percent in paroxetine patients compared with 10 percent in fluvoxamine patents (P=0.028).
The third trial assessed differences in adverse events between fluvoxamine (100–150 mg/d) and fluoxetine (20–80 mg/d) in 100 patients over 7 weeks.67 Fluoxetine-treated patients suffered under nausea significantly more often than fluvoxamine patients (42.5% compared with NR; P=0.03)
A fair-rated, Dutch prospective observational study followed 1,251 patients for up to 12 months to assess adverse events of sertraline (N=659) compared to other SSRIs (fluoxetine, fluvoxamine, paroxetine).230 No exclusion criteria were applied. Psychiatrists recorded adverse events at each patient visit. The WHO adverse reaction terminology was used for outcome assessment. Significantly more sertraline patients had the diagnosis of depressive disorder at baseline (P<0.001). Overall, 74.1 percent of patients reported at least one adverse event. Diarrhea occurred more frequently in the sertraline group than in the other SSRI groups (P<0.05). However, abdominal pain was reported more frequently by other SSRI users than sertraline users (P<0.05). No other adverse event differed significantly across groups.
We pooled data from efficacy trials to assess differences in overall loss discontinuation rates, discontinuation rates because of adverse events, and discontinuation rates because of lack of efficacy of SSRIs as a class compared to other second-generation antidepressants in adult outpatients with MDD (Exhibit 6). Available data were insufficient to determine some results for desvenlafaxine and nefazodone. The only statistically significant difference in pooled estimates was a higher discontinuation rate because of adverse events for venlafaxine-treated patients than for patients on SSRIs (RR, 1.42; 95% CI 1.16 to 1.73). Overall, this finding was balanced by lower discontinuation rates because of lack of efficacy for venlafaxine (RR, 0.75; 95% CI 0.53 to 1.05). No significant differences could be detected between SSRIs and mirtazapine or between SSRIs and bupropion. Numerical differences in discontinuation rates attributed to adverse events generally favored SSRIs over mirtazapine and bupropion but did not reach statistical significance.
A meta-analysis of 15 RCTs did not find any statistically significant differences in discontinuation rates because of adverse events between fluoxetine and other SSRIs as a class.231
B. Specific Adverse Events
A nested case control study examined the risk of sudden cardiac death or near death in patients treated with citalopram, fluoxetine, or venlafaxine.232. The study was based on the United Kingdom General Practice Research Database which included data on more than 207,000 patients who initiated treatment with citalopram, fluoxetine, or venlafaxine for MDD or anxiety. The follow-up time was an average of 3.3 years. Within the cohort, 568 cases of sudden cardiac arrest or near death occurred. These cases were matched with more than 14,000 controls. Results showed that no significant differences in risks for sudden cardiac death or near death were obvious between the examined medications. The adjusted odds ratio associated with venlafaxine relative to fluoxetine was 0.66 (95% CI 0.38–1.14), of venlafaxine relative to citalopram was 0.89 (95% CI 0.50–1.60).
A well conducted Dutch study by Trifirò et al investigated the association between ischemic stroke and SSRIs in 996 Dutch patients, 65 years and older, included in a longitudinal general practice research database (Integrated Primary Care Information Database). Results of this population-based, nested case-control study showed a significantly increased risk of stroke with respect to the current use of SSRIs compared with non-use (OR 1.55; 95% CI 1.07–2.25), particularly when antidepressants were used for less than six months. No excess risk could be found for the use of tricyclic and other antidepressant drugs.
Another good, nested case-control study conducted in patients on antidepressant medication included in an American multi-state managed care organization medical claims database found similar results. 233 The risk of ischemic stroke in current SSRI users compared with remote or nonusers was significantly increased. (adj. HR:1.55; 95% CI 1.00–2.39), whereas the risk of hemorrhagic stroke in current users of SSRIs was not significantly different compared to that of remote or nonusers. (adj. HR: 1.18; 95% CI 0.64–2.16)
Likewise, a fair case-control study including 916 cases of intracerebral or subarachnoid hemorrhage did also not detect any association between hemorrhagic stroke and SSRIs (OR: 1.1; 95% CI 0.7–1.8; P=0.63)235.
A fair case-control study 236 evaluated the risk of idiopathic venous thromboembolism in 782 patients aged 70 years or younger with a first time diagnosis of venous thromboembolism and concurrent use of antidepressant medication. The study, which included SSRIs, tricyclic antidepressants and other antidepressants, found no increased risk of idiopathic venous thromboembolism among users of SSRIs. The unadjusted OR for current use of SSRIs compared with nonusers of any antidepressant (past use and nonusers combined) was 0.9 (95% CI 0.6–1.2).
Changes in weight
A 32-week acute and continuation trial assessed differences in weight changes among patients treated with fluoxetine, paroxetine, and sertraline.124 Paroxetine patients showed a significantly greater mean weight change (+3.6%) than did those taking fluoxetine (−0.2%; P=0.015) and sertraline (+1.0%; P<0.001). Significantly more patients in the paroxetine group (25.5%) had a weight gain of more than 7 percent than in the fluoxetine (6.8%; P=0.016) and sertraline groups (4.2%; P=0.003). A 1-year, placebo-controlled continuation trial of fluoxetine reported similar findings.73 Initially, fluoxetine treatment led to a modest weight loss; from week 12 to week 50, however, a significant weight gain compared to placebo was reported (+3.1 kg; P<0.001). An open-label, nonrandomized, 2.5-year study on OCD patients also reported the lowest increase in weight gain for fluoxetine (+0.5 kg). Other SSRIs lead to greater weight gains (sertraline +1.0 kg; citalopram +1.5 kg; paroxetine +1.7 kg; fluvoxamine +1.7 kg), however, differences are neither statistically nor clinically significant.237 A pooled analysis of two RCTs comparing escitalopram and paroxetine reported a similar gain in body weight for both patient groups.238 After 27 weeks of followup, patients on escitalopram gained 1.68 kg and patients on paroxetine gained 1.64 kg.
A double-blinded placebo-controlled 52-week acute and continuation trial assessed weight changes during bupropion treatment.239 Bupropion-treated patients showed a modest but nevertheless significant decrease of body weight from baseline (−1.15 kg; P<0.001). The magnitude of weight change was closely related to the body mass index (BMI). Patients with a higher BMI experienced greater weight loss.
Consistently, studies comparing mirtazapine with other second-generation antidepressants reported higher weight gains for mirtazapine than for the comparator groups. In three RCTs, these differences reached statistical significance.88–90 Mean weight gains ranged from 0.8 kg to 3.0 kg after 6 to 8 weeks of treatment.
Evidence from one good240 and two fair case-control studies241, 242 indicate an increased risk of upper gastrointestinal tract bleeding during SSRI treatment. The good quality case control study matched 11,025 case patients suffering from bleeding abnormalities with 21,846 control patients. In addition, the study compared 1,008 patients with gastrointestinal bleeding with 1,990 control patients based on the ARNO database, a population-based database for drug use in Italy. This study excluded patients with a prescription for non-steroidal anti-inflammatory drugs (NSAIDs), corticosteroids, antihemorrhagics and antithrombotic agents.
Seven percent of case patients with any bleeding disorder and 6.9 percent of control patients, as well as 8.6 percent of case patients with upper gastrointestinal bleeding and 6.3 percent of control patients were on antidepressants (SSRIs, TCAs, and other antidepressants). None of the studied antidepressants of interest (citalopram, escitalopram, fluoxetine, fluvoxamine, mirtazapine, paroxetine, sertraline, and venlafaxine) were individually associated with an increased risk for either bleeding abnormalities or gastrointestinal bleeding. Furthermore, SSRIs as a class also did not yield an increased risk of any bleeding abnormality (OR 0.99; 95% CI 0.89 – 1.10). With respect to gastrointestinal bleeding, SSRIs as a class exhibited a numerically increased risk that did not reach statistical significance (OR 1.31; 95% CI 0.91 – 1.88). 240
The other two included studies confirm an increased risk for upper gastrointestinal bleeding for patients on second-generation antidepressants.241, 242 In contrast to the Italian-based study reported above, the two studies also enrolled patients who were on NSAIDs and other drugs.
One study matched 1,552 case subjects with 68,590 control subjects using the Manitoba Population Health Research Data Repository. SSRIs were associated with a statistically significant increase in the risk of upper gastrointestinal bleeding (adjusted OR 1.43; 95% CI 1.09 – 1.89) 241. Furthermore, this study investigated the effect of the combination of different drugs with SSRIs. The risk of suffering from upper gastrointestinal bleeding was higher in case subjects being medicated with SSRIs and non-steroidal anti-inflammatory drugs (NSAID) (OR 3.17; 95% CI 2.01 – 5.00). Proton pump inhibitors had a protective effect (albeit not statistically significant) on upper gastrointestinal bleeding in patients on SSRIs (OR 0.56; 95% CI 0.24 – 1.30).
The other case control study was based on data from the Health Improvement Network database in the United Kingdom and provided similar findings. The study revealed a statistically significant association between a higher risk of upper gastrointestinal tract bleeding and the use of SSRIs (OR1.6; 95% CI 1.2 – 2.1) as well as SNRIs (OR 2.9; 95% CI 1.5 – 5.6). 242
We identified two studies assessing the risk of fractures for subjects on antidepressant medication. 243, 244. Both studies reported an increased fracture risk for patients with antidepressant intake. The larger study, a well conducted case-control study including 498,617 subjects (124,655 cases and 373,962 controls) from a Danish national prescription database, reported a significant dose-response relationship for citalopram, fluoxetine and sertraline with respect to an increase of the risk of fracture. 243 Amongst SSRIs, high-dose citalopram, fluoxetine, paroxetine, and sertraline were associated with the highest risk for hip fracture (OR 1.98, 95% 1.82–2.16) and other fractures except fractures of the forearm and spine (OR 1.38, 95% CI 1.33–1.44). Evidence regarding the impact of the duration of use on the risk of fractures was mixed for second-generation antidepressants.
Findings of the Danish cohort study described above were consistent with results of a fair, population - based, prospective cohort study on the risk of nonvertebral fractures during antidepressant treatment.244 This study on 7983 Dutch men and women, aged 55 years or older, revealed a 2.35 times higher risk of nonvertebral fracture for current users of SSRIs compared with non-users of antidepressants. (95% CI, 1.32–4.18). Subjects, who had been using SSRIs for at least six months had a 3.36 fold higher risk of fractures (95% CI, 1.39–8.08).
Evidence from controlled trials and observational studies is also insufficient to conclude for or against an increased risk of liver toxicity during nefazodone treatment. Nevertheless, numerous case reports not included in this report contain low-level quality but potentially important evidence citing an increased risk of liver toxicity during nefazodone treatment.245 One maker of nefazodone has announced that it is withdrawing the drug from the US market by June 2004 because of safety concerns (websource: www.medscape.com/viewarticle/47852; accessed 5-20-2004). An analysis of AERS data and a claims database on more than 60,000 patients who initiated duloxetine or venlafaxine found no difference in the risk of hepatic injury between the two drugs.246
Evidence from controlled trials and observational studies is insufficient to conclude for or against an increased risk of hyponatremia in patients treated with SSRIs. However, the methods of our report did not include case reports and case series. The published literature includes numerous case reports of hyponatremia and inappropriate secretion of antidiuretic hormone as rare side effects.247 Even if this evidence is considered weak, it could be important in the absence of studies with the methodological strength to account for rare adverse events.
Evidence from controlled trials and observational studies is insufficient to conclude for or against an increased risk of seizures in patients taking any of the reviewed drugs, including bupropion. An analysis of FDA data derived from approval reports indicated a higher risk of seizures for bupropion compared with other antidepressants.248 Overall, 0.6 percent of patients treated with bupropion experienced seizures. The standardized incidence ratio compared with placebo was 1.58 (1.03, 2.32).
A recent chart review of 538 patients with deliberate self-poisoning with antidepressants reported that seizures were more common in patients with venlafaxine overdose than in patients with TCA or SSRI overdose.249
A subgroup analysis of a good Swedish RCT examined the incidence of sexual side effects from citalopram (20–60 mg/d) compared to those from sertraline (50–150 mg/d)53, 250 in 308 study completers with MDD. Outcome assessment was conducted at baseline and at week 24. Citalopram and sertraline did not differ significantly in the magnitude and frequency of sexual side effects. Only one patient was lost to follow-up attributable to sexual side effects in this study. Similarly, citalopram did not differ from paroxetine in sexual side effects in a nonrandomized trial.251
A good meta-analysis including data on 1,332 patients reported a significantly higher rate of sexual satisfaction in bupropion- than in SSRI-treated patients with MDD (RR 1.28; 95% CI 1.16 to 1.41).108
Multiple studies indicated that bupropion has a lower risk of sexual dysfunction than some SSRIs.110, 114, 115, 128, 252 Three studies assessed the incidence of sexual dysfunction in depressed outpatients treated with bupropion or sertraline.114, 115, 128
Two fair-rated RCTs compared the incidence of sexual dysfunction in 360 and 364 patients with MDD during 8 weeks of treatment with bupropion (150–400 mg/d), sertraline (50–200 mg/d), or placebo.114, 115 Outcome measures were efficacy (HAM-D, CGI) and sexual dysfunction as assessed by investigators using DSM-IV definitions for sexual dysfunction disorders. Intention-to-treat analyses yielded no significant differences between bupropion and sertraline in any efficacy measures at trial endpoints. During the studies, sertraline showed more sexual adverse events than bupropion at various time points. However, in one trial overall satisfaction with sexual function did not differ significantly between the bupropion and the sertraline group at endpoint.114 In the other study, beginning at day 21 until the end of the study, the overall satisfaction with sexual function was significantly higher in the bupropion group than in the sertraline group (P<0.05).115
The third RCT assessed the sexual side effects of bupropion SR (150–400 mg/d) and sertraline (100–300 mg/d) in 248 depressed outpatients.128 Study duration was 16 weeks; loss to follow-up was 31.5 percent. Sexual dysfunction was determined by investigator interviews and patient-completed questionnaires. Treatment groups were comparable at baseline. Intention-to-treat analysis showed that, beginning at day 7, significantly fewer bupropion-treated patients than sertraline-treated patients reported sexual dysfunction (P<0.001) throughout the study. These findings were significant for males (P<0.05) and for females (P<0.01). Significantly more patients in the sertraline group developed sexual arousal disorder, orgasm dysfunction, or ejaculation disorder (men: 63% compared with 15%; P<0.001; women: 41% compared with 7%; P<0.001).
The combined NNT to yield one additional person who is satisfied with the overall sexual function is 7.
A fair, 8-week RCT compared efficacy and sexual side effects of bupropion (150–400 mg/d), fluoxetine (20–60 mg/d), and placebo in 456 outpatients with MDD.110 Loss to follow-up was 36 percent. Efficacy did not differ significantly. Bupropion had more remitters than fluoxetine (47% compared with 40%) at endpoint. Bupropion also showed significantly fewer sexual side effects than fluoxetine throughout the study. Beginning at week 1 until endpoint, significantly more fluoxetine-treated patients were dissatisfied with their overall sexual function than bupropion-treated patients (P<0.05).
Similarly, a fair 8-week RCT comparing bupropion with paroxetine reported significantly lower rates of sexual dysfunction for bupropion than for paroxetine (Sex Effects Scale, P<0.05).253 Subgroup analysis revealed that a significant difference in anti-depressant related sexual dysfunction was detected in men but not in women.
The largest observational study was a Spanish open-label, prospective study using the Psychotropic-Related Sexual Dysfunction Questionnaire (PRSexDQ) in 1,022 outpatients treated with various antidepressants.254 All patients had normal sexual functioning at study onset. Overall, 59 percent of patients experienced some type of sexual dysfunction. Among second-generation antidepressants, citalopram, paroxetine, and venlafaxine had the highest incidence of sexual dysfunction (73 percent, 71 percent, and 67 percent, respectively); mirtazapine and nefazodone had the lowest (24 percent and 8 percent, respectively). This study did not include data on bupropion and escitalopram. In another observational study, findings of a cross-sectional survey of patients on second-generation antidepressants presented similar results.255 Paroxetine had the highest rate of sexual dysfunction; nefazodone and bupropion had the lowest.
Sexual side effects were also commonly reported adverse event for SSRIs and SNRIs in efficacy trials. Most of these studies did not report the use of targeted questions for sexual side effects. Therefore, patient-reported numbers might not reflect the true incidence. Paroxetine- and sertraline-treated patients frequently reported significantly higher rates of sexual side effects72, 82, 83, 91, 113, 121 than did patients in the active control groups. In one trial, significantly more patients on sertraline withdrew because of sexual side effects than did patients on bupropion (3.3% compared with 13.5%; P=0.004).113 In another study patients on duloxetine reported statistically significantly lower rates of sexual dysfunction than patients on escitalopram (33% compared with 49%; P=0.01).256
In 2004 an Expert Working Group of the UK Committee on Safety in Medicines (CSM) investigated ongoing safety concerns about suicidal behavior with some second-generation antidepressants (citalopram, escitalopram, fluoxetine, fluvoxamine, mirtazapine, paroxetine, sertraline, venlafaxine) in patients with MDD.257 The Expert Working Group studied data from 477 published and unpublished randomized controlled trials on more than 40,000 individuals. However, these data were limited to studies funded by the pharmaceutical industry.
In summary, the Expert Group advised that the balance of risks and benefits for the treatment of depression in children less than 18 years is unfavorable for citalopram, escitalopram, mirtazapine, paroxetine, sertraline, and venlafaxine. Only fluoxetine appeared to have a favorable risk-benefit ratio. Fluvoxamine could not be assessed for pediatric use because of lack of data. Conclusions were based on the fact that, with the exception of fluoxetine, clinical trial data failed to demonstrate efficacy in a pediatric population. In addition, an increased risk of suicidal thoughts and self-harm was observed consistently across drugs.
For adults, clinical trial data consistently showed that the risk of suicide-related events in patients receiving second-generation antidepressants is higher than in patients on placebo. However, none of the pooled estimates for individual drugs reached statistical significance. The risk of suicide-related events was similar between second-generation antidepressants and active comparators.
A meta-analysis limited the CSM data to placebo-controlled trials of SSRIs in adults. Results did not yield any evidence that SSRIs increase or protect against the risk of suicide (OR 0.85; 95% CI 0.20 to 3.40).258 However, weak evidence of an increased risk of self-harm was detected (OR 1.57; 95% CI 0.99 to 2.55).
In addition, the Expert Group commissioned an observational study (a nested case-control study) using the General Practice Research Database (GPRD) to investigate the association between antidepressants and self-harm based on data on more than 146,000 patients with a first prescription of an antidepressant for depression.259 This study did not find any evidence that the risk of suicide (OR 0.57; 95% CI 0.26 to 1.25) or self-harm (OR 0.99; 95% CI 0.86 to 1.14) is greater in patients on second-generation antidepressants than in patients on TCAs.
Findings of other studies are mixed.152, 260–277 A good meta-analysis of published data on more than 87,000 patients in SSRI trials for various conditions reported a significantly higher risk of suicide attempts for SSRI patients than for placebo-treated patients (2.25; 95% CI 1.14 to 4.55).278 Furthermore, an increase in the odds ratio of suicide attempts was observed for SSRIs compared to interventions other than TCAs (OR 1.94; 95% CI 1.06 to 3.57). No significant difference existed in the pooled analysis of SSRIs compared to TCAs (OR 0.88; 95% CI 0.54 to 1.42).
A fair-rated open cohort study using UK data observed 172,598 people to compare the suicide rates of 10 commonly used antidepressants (fluoxetine, dothiepin, amitriptyline, clomipramine, imipramine, flupenthixol, lofepramine, mianserin, doxepin, and trazodone) for 5 years.266 Suicide was the main outcome measure. Dothiepin was the most commonly prescribed antidepressant and was used as a reference drug. Compared with dothiepin, only fluoxetine (RR 2.1; 95% CI 1.1 to 4.1) and mianserin (RR 1.8; 95% CI 1.0 to 3.6) yielded a significantly higher relative risk for suicide. Relative risks did not differ among patients who had no history of being suicidal and had been prescribed only one antidepressant. A recent matched case-control study using data of 159,810 patients in the UK did not support these findings.279 A total of 555 cases of nonfatal suicidal behavior were matched with 2,062 controls. Compared to dothiepin, the risk of suicidal behavior was similar among users of amitryptilin (RR: 0.83; 95% CI 0.61 to 1.13), fluoxetine (RR 1.16; 95% CI 0.90 to 1.50), and paroxetine (RR: 1.29; 95% CI 0.97 to 1.70).
A retrospective review of data in FDA summary reports compared the absolute suicide rate and the suicide rate by patient exposure-years of SSRIs (citalopram, fluoxetine, fluvoxamine, paroxetine, sertraline), other antidepressants (nefazodone, mirtazapine, bupropion, maprotiline, trazodone, mianserin, dothiepin, imipramine, amitriptyline, venlafaxine), and placebo.268 Crude suicide rates and adjusted suicide rates did not differ significantly by patient exposure-years among patients assigned to SSRIs, other antidepressants, or placebo. A Spanish database review did not find significant differences in suicidal ideation between paroxetine, imipramine, amitriptylyne, clomipramine, mianserin, doxepin, maprotiline and placebo.269 A retrospective cohort and a nested case control study using data from a New Zealand database reported a higher rate of self-harms in SSRI- than in TCA-treated patients (OR: 1.66; 95% CI 1.23 to 2.23) but no differences in suicides.280 However, no differences in self-harm or suicides were apparent among citalopram-, fluoxetine-, or paroxetine-treated patients.
Findings of the CSM Expert Group on suicidality in children are consistent with results from an earlier NICE (National Institute for Clinical Excellence) report.146 In patients younger than 18 years the risk of self-harm was significantly greater in patients on SSRIs than on TCAs (OR 1.59; 95% CI 1.01 to 2.50). Although no statistically significant differences among SSRIs were detected, the greatest risk of self-harm was among paroxetine users. A retrospective cohort study on almost 21,000 children who had initiated antidepressants 281 and an analysis of FDA data282 reported similar results. The use of antidepressant drugs in pediatric patients was associated with statistically significant increase in suicidality (RR: 1.66; 95% CI 1.02 to 2.68). The rate of suicidal event was 27.04 per 1000 patient years for children, compared with an event rate of 4.4 to 9.1 suicidal events per 1000 patient years in adult populations. 277, 281
Results of other studies are mixed.283–285 Two studies reported that second-generation antidepressants increase the risk of suicidality in adolescents but decrease the risk in adults 273, 274 The first study, a meta-analysis of observational studies in a combined population of more than 200,000 patients indicated that the use of SSRIs significantly increase the risk of attempted or completed suicides in adolescents (OR 1.92; 95% CI 1.51–2.44). The risk of attempted or completed suicide among adults, however, was significantly decreased in adults (OR 0.57,95% CI 0.47–0.70) and among people aged 65 years or older (OR 0.46, 95% CI 0.27–0.79). 274 These findings are consistent with a case-control study of more than 1000 adolescents and adults treated with antidepressants for MDD273 and an unpublished FDA data-analysis on more than 99,000 participants of 372 trials.286 The FDA pointed out that the risk of suicidality is increased in children and patients 18 to 24 years but not in other adult patients.
Other adverse events
A database analysis in the UK on fatal toxicity of second-generation antidepressants found venlafaxine to have the highest fatal toxicity rate (13.2/1,000,000 prescription) among second-generation antidepressants.287
A case-control study did not find an association between SSRIs and breast cancer.215 Evidence from randomized trials and observational studies is insufficient to draw conclusions regarding the risk of rare but potentially fatal adverse events such as hyponatremia or liver toxicity. However, multiple case reports have indicated that many of the SSRIs are associated with hyponatremia, especially in older patients.247 Similarly, reports of liver toxicity with nefazodone have not been confirmed by controlled trials and observational studies.245 Owing to a lack of studies with the methodological strength to assess these rare events, conclusions should be made on other grounds such as comorbidities, taking case reports into consideration.
A case control study based on a cohort of 165,958 patients with depression included in the UK General Practice Research Database, selected a total of 2,243 cases of incident diabetes mellitus and 8,963 matched comparison subjects.288 Results showed that recent long-term use (> 24 months) of antidepressants in moderate to high daily doses was associated with an increased risk of diabetes (incidence rate ratio, 1.84; 95% CI, 1.35–2.52). For users of SSRIs as a group, increased risk was observed only for recent long-term use of moderate to high daily doses (incidence risk ratio, 2.06; 95% CI, 1.20–3.52). When individual antidepressants were analyzed, increased risk estimates only in long-term users were observed for recent use of fluvoxamine, paroxetine and venlafaxine.
C. Summary of the evidence
Fair to good evidence from multiple randomized controlled head-to-head trials and retrospective data analyses of prescription event monitoring documents that adverse events profiles are similar among reviewed drugs. Frequencies of some adverse events, however, differ among drugs. Venlafaxine had a significantly higher rate of nausea and vomiting in multiple trials; paroxetine frequently led to higher sexual side effects; mirtazapine to higher weight gains; and sertraline to a higher rate of diarrhea than comparable second-generation antidepressants. A retrospective review of prescription event monitoring data provides fair evidence that, among SSRIs, fluvoxamine has the highest mean incidence of adverse events.227 Pooled estimates from efficacy trials suggest that venlafaxine has a statistically significantly higher rate of discontinuation because of adverse events than do SSRIs as a class. However, overall discontinuation rates do not differ significantly between venlafaxine and SSRIs.
Cardiovascular adverse events
Fair evidence from one case-control study with 568 cases of sudden cardiac death or near death revealed no significant differences in risk among citalopram, fluoxetine, or venlafaxine 232. Evidence from two well conducted case-control studies, each including about 1000 cases, indicates that the use of SSRIs leads to a significantly increased risk of ischemic stroke compared to non-use.233, 234 No association, however, between SSRIs and an increased risk for hemorrhagic stroke could be detected.233, 235
A fair rated case-control study reported no increased risk of idiopathic venous thromboembolism among users of SSRIs.236
Evidence from a well conducted case control study including 124,655 cases indicates a dose-response relationship for citalopram, fluoxetine, paroxetine, and sertraline for risk of fracture.243 Results of a fair rated prospective cohort study including individuals aged 55 and older, indicate an increased risk of nonvertebral fractures for current users of SSRIs compared with nonusers. (HR: 2.35, 95% CI 1.32–4.18).244
Fair to good evidence from three case control studies indicate an increased risk of upper gastrointestinal tract bleeding during SSRI and SNRI treatment.240–242 The combination of SSRIs and NSAIDs appears to further increase the risk of gastrointestinal bleeding.
Eight trials and a pooled analysis of two identical RCTs provide evidence that bupropion causes lower rates of sexual dysfunction than escitalopram,42 fluoxetine 109, 110 paroxetine,253 and sertraline.110, 115, 128 The NNT to yield one additional person with a high overall satisfaction of sexual functioning is 7. This treatment effect was consistent across all studies. A cross-sectional survey supports this evidence by reporting the lowest rates of sexual side effects for bupropion and nefazodone in patients treated with SSRIs or other second-generation antidepressants.255 Multiple trials give fair evidence that paroxetine, sertraline, and mirtazapine tend to have higher rates of sexual side effects than other second-generation antidepressants.72, 73, 82, 83, 91, 113, 121, 255
Evidence from controlled trials and large observational studies is mixed about a higher risk of suicidality in patients treated with second-generation antidepressants. Large observational studies suggest that the risk is significantly increased in children, adolescents and young adults but not in older adult patients.273, 274 Current data does not suggest any differences in risks among second-generation antidepressants.146, 277, 282
Key Question 3. Are there subgroups of patients based on demographics (age, racial groups, sex), other medications, or co-morbidities for which one second-generation antidepressant is more effective or associated with fewer adverse events?
We did not find any studies directly comparing the efficacy, effectiveness, and tolerability of second-generation antidepressants between subgroups and the general population. However, several studies conducted subgroup analyses or used subgroups as the primary study population. Results can provide indirect evidence for Key Question 3. Included studies are presented in Table 22
We did not include any placebo-controlled studies on this topic as there were ample head-to-head trials.
Citalopram compared with sertraline
One randomized trial evaluated citalopram and sertraline in the treatment of 138 non-demented elderly patients with minor depressive disorder and subsyndromal symptomatology.137 Although this trial does not meet our eligibility criteria because of the study design (nonrandomized trial), we are briefly summarizing it because it is the only evidence pertaining to a comparison of these two SSRIs. Both treatments improved depressive symptoms (as measured by the HAM-D scale); HAM-D remission rates were similar for citalopram and sertraline at the end of the study (53% and 42%, P=0.25). Similar improvements were seen in Global Assessment of Function (GAF) and cognitive scores.
Escitalopram compared with fluoxetine
One 8-week study compared escitalopram, fluoxetine, and placebo in 518 participants older than 65 years of age (mean age in each treatment group, 75 years).65 Outcome measures included the MADRS and the CGI-S. Patients on escitalopram experienced greater improvement than those on fluoxetine in MADRS score (using LOCF analysis) at week 8 (P < 0.01); however, the patients treated with escitalopram and with placebo did not differ significantly. Escitalopram, placebo, and fluoxetine MADRS response rates were similar (46%, 47%, and 37%, respectively, P=not significant). In addition, MADRS remission rates were similar for escitalopram and placebo (40% and 42%), but for fluoxetine compared with placebo, the difference was statistically significant (30% compared with 42%, P=0.05). Escitalopram- and fluoxetine-treated patients experienced significantly more nausea than placebo-treated patients (P < 0.01).
Fluoxetine compared with paroxetine
Two RCTs were conducted in a population older than 60 years.68, 71 The first trial was an Italian study lasting 1 year that enrolled 242 patients to determine the effects of fluoxetine (20–60 mg/d) and paroxetine (20–40 mg/d) on mood and cognitive function in depressed, non-demented persons (65 years or older). Both groups significantly improved on their HAM-D scores and cognitive performance. Paroxetine showed a faster onset of action and a significantly greater improvement of HAM-D scores during the first 6 weeks (Week 3: P<0.05; Week 6: P<0.002). A Kaplan-Meier analysis evaluating the percentage of responders over time revealed a significant difference in favor of paroxetine (P<0.002). Treatment groups did not differ significantly in CGI scores. Fluoxetine had a significantly greater number of patients with severe adverse events than paroxetine (22 compared with 9; P<0.002). However, loss to follow-up in this study was 39.3 percent, so the validity of the results should be viewed cautiously.
The second trial conducted in an elderly population enrolled 108 patients with major depression in Austria and Germany for 6 weeks using the same dosage as the Italian study.71 Loss to follow-up was not reported. An intention-to-treat analysis revealed no differences between the treatment groups in changes of scores on MADRS and HAM-D; the paroxetine group had significantly more responders at 6 weeks on MADRS and HAM-D scales (37.5%compared with 17.5%; P=0.04). Patients on paroxetine also had significantly better MMSE and SCAG scores assessing cognitive function at Week 3 than did those on fluoxetine. No statistically significant differences in adverse events were reported.
Fluoxetine compared with sertraline
One fair, 12-week study comparing fluoxetine to sertraline was conducted in 236 participants older than 60 years.77, 79 Loss to follow-up was 32.2 percent. In this study, outcome measures also included quality of life (Q-LES-Q) and cognitive assessments (SLT, MMSE, Digital Symbol Substitution Test). Fluoxetine- and sertraline-treated patients did not differ significantly on primary outcome measures (MADRS, HAM-D). Response rates (fluoxetine, 71%; sertraline, 73%) and remission rates (46% compared with 45%) were similar. Quality of life and other patient-rated secondary efficacy measures were similar for both treatment groups at endpoint. Sertraline-treated patients showed a greater cognitive improvement on the Digit Symbol Substitution Test at endpoint (P=0.037). A subgroup analysis of 75 patients 70 years of age or older showed a greater response rate for sertraline-treated patients (P = 0.027).79
A subgroup analysis of a long-term effectiveness trial comparing fluoxetine, paroxetine, and sertraline reports similar response and remission rates for patients older than 65 years and the general study population.55
Mirtazapine compared with paroxetine
A fair trial randomized 255 elderly participants for eight weeks.89 Loss to follow-up was 27 percent. Mirtazapine and paroxetine were equally effective in reducing HAM-D scores at the endpoint, but mirtazapine lead to a faster response. A Kaplan-Meier analysis showed a significantly faster time to response for mirtazapine (mean 26 days compared with mean 40 days for paroxetine; P=0.016). No significant difference in response rates on the CGI scale was noted. Significantly more mirtazapine-treated patients reported weight gain (P<0.05). Paroxetine-treated patients reported a significantly higher rate of nausea, tremor, and flatulence (P<0.05).
Venlafaxine compared with citalopram
A fair European 6-month study compared venlafaxine ER (37.5–150 mg/d) to citalopram (10–30 mg/d) for the treatment of depression in elderly outpatients (mean age 73 years).92 No statistical differences in any outcome measures (MADRS< CGI-S, CGI-I) could be detected at study endpoint. The remission rates were 19 percent for venlafaxine and 23 percent for citalopram. Both treatment groups reached a 93 percent response rate.
Venlafaxine compared with fluoxetine
One fair trial compared venlafaxine IR (37.5 – 225 mg/d) to fluoxetine (20 – 60 mg/d) for the treatment of unipolar depression in elderly patients (mean age 71 years).46 Both treatment groups experienced a significant reduction in HAM-D total scores at 8 weeks; however, there were no significant differences between groups in HAM-D, MADRS, or CGI scores at endpoint. Remission rates at 8 weeks were 27 percent for venlafaxine and 20 percent for fluoxetine. Venlafaxine-treated patients experienced significantly higher rates of nausea (45% compared with 23%), dry mouth (23% compared with 6%) and constipation (22% compared with 10%); P<0.01 for all three comparisons.
Venlafaxine compared with sertraline
One study determined efficacy and safety of venlafaxine (25–100 mg/d) compared to sertraline (18.5–150 mg/d) in 52 frail nursing home residents (61 to 99 years of age).293 We graded the quality of this study as poor for efficacy because of high loss to follow-up (44.2%), but we note it here because it is the only study comparing these two agents, and because the high loss to follow-up may be expected in this population (elderly nursing home residents). The investigators reported a significantly higher rate of withdrawal among venlafaxine- than sertraline-treated patients (63% compared with 24%). In addition, venlafaxine-treated patients had a significantly higher rate of severe adverse events (P=0.022) and withdrawal because of severe adverse events or side effects (P=0.005) than did the sertraline-treated patients.
Venlafaxine compared with SSRIs
A pooled data analysis combined original data from eight comparable, double-blind, active-controlled, randomized trials.294, 295 A primary objective of this analysis was to determine differences in response and remission based on sex and age. This study was not based on a systematic literature search, so results must be viewed cautiously. For venlafaxine-treated patients, neither age (< 50 or ≥ 50 years of age) nor sex affected remission rates.295 Among patients treated with SSRIs, however, a significant interaction was observed between treatment and sex (P=0.004); older women had a poorer SSRI response (response rate: 28%) than younger women (response rate: 36%), and both older and younger men (response rates: 35% and 36%, respectively). Remission rates for older women treated with venlafaxine (48%) were higher than remission rates for older women treated with SSRIs (28%, P=0.0004). Hormone replacement therapy appeared to eliminate these differences. Additional analyses of age subgroups (≤ 40, 41–54, 55–64, and ≥ 65 years of age) and sex subgroups revealed that no significant age-by-treatment, sex-by-treatment, or age-by-sex-by-treatment interactions occurred. Men and women of different ages within each treatment group had similar rates of remission, response, and absence of depressed mood.294 Among patients over 40 years of age, the rates of adverse events were similar between the treatment groups, although venlafaxine-treated patients aged 55 to 64 years reported significantly more nausea than placebo (P<0.003), and placebo patients aged 41 to 54 years reported a significantly higher frequency of headaches than venlafaxine (P<0.01).
Bupropion compared with paroxetine
One fair RCT examined the efficacy of bupropion SR (100–300 mg/d) and paroxetine (10–40 mg/d) in 100 outpatients ages 60 years or older (range 60–88 years) over 6 weeks.111, 112 The majority of patients were white (bupropion SR, 98%; paroxetine, 90%), female (bupropion SR, 54%; paroxetine, 60%), and did not use antidepressants for the current episode before enrollment (bupropion SR, 83%; paroxetine, 88%). Statistical analysis used a LOCF method. The overall loss to follow-up was 16 percent with no significant difference between treatment groups. Efficacy according to any outcome measure did not differ significantly between treatment groups. Response rates (≥ 50% reduction in HAM-D scores) were similar in both groups (bupropion SR, 71%; paroxetine, 77%). Quality-of-life scales (QLDS, SF-36) showed statistically significant improvements in both treatment groups from baseline to endpoint (P<0.0001), but they did not differ significantly between treatment groups.
No studies directly compared the efficacy, effectiveness and harms of second-generation antidepressants among different races or ethnicities. Therefore, we summarize results of studies that compared second-generation antidepressants with placebo.
Duloxetine compared with placebo
Two pooled analyses of seven placebo-controlled duloxetine trials assessed the efficacy and tolerability of duloxetine in Hispanic296 and African American patients297 compared to Caucasian patients. The first analysis included 1,342 Caucasians and 120 Hispanics and found no difference in efficacy outcomes for Hispanics and Caucasians.296 There were no significant differences between groups in discontinuation rates due to adverse events ir in the types or occurrence of specific adverse events. The second analysis of 1,300 Caucasians and 123 African Americans also found no evidence for a differential effect of duloxetine in African-American and Caucasian patients in efficacy or safety outcomes.297
Fluoxetine compared with placebo
An RCT examined ethnic differences in response to antidepressant treatment among depressed HIV-positive patients.298 A total of 118 patients were randomized to either fluoxetine (20–80 mg/d) or placebo for 8 weeks. Of all participants, 67 percent were White, 19 percent Black, and 14 percent Latino; only 1.1 percent (N=2) were female. Loss to follow-up was significantly greater among Latinos (53%) than among Blacks (14%) and Whites (28%; P<0.05). Ethnicity was not associated with the total number of treatment emergent side effects or dosage. Among completers within the active-treatment group, Whites were more likely to respond to treatment than the other two groups (84% compared with 50% in Blacks and 67% in Latinos). Among completers in the placebo group, Latinos were more likely to show treatment response (80%) than were blacks (36%) or whites (43%). However, a statistical analysis of these findings was not possible because of the low number of Latinos who completed the study.
Paroxetine compared with placebo
A pooled analysis of 104 paroxetine trials (14,875 patients) detected slightly lower response rates for Hispanics and Asians than for Blacks and Whites.299
One study that did not meet our inclusion criteria performed a secondary analysis of data from the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study to compare remission and response rates among Blacks, Whites, and Hispanics with nonpsychotic MDD.300 We briefly describe it here because because of the paucity of evidence on this topic. STAR*D included outpatients in 23 psychiatric and 18 primary care centers. Participants received flexible doses of citalopram for up to 14 weeks. There were significant differences in baseline characteristics among ethnic groups. Prior to adjustment for such differences, Black participants had lower HRSD17 remission rates (18.6%) than white (30.1%) or Hispanic participants participants (24.2%). After adjustments, there were no significant differences in HRSD remission rates among groups; however, remission rates were still lower for Blacks compared to whites based on the Quick Inventory of Depressive Symptomatology-Self Report (QIDS-SR). In general, Black and Hispanic participants had poorer responses to citalopram compared to White participants.
A pooled data analysis of venlafaxine and SSRIs described above294, 295 did not find any significant associations between sex and outcomes or sex and treatment of MDD. Among patients treated with SSRIs, however, a significant interaction was observed between treatment and sex (P=0.04); older women had a poorer SSRI response (28%) than younger women (36%) and than both, older and younger men (35% and 36%, respectively). Additional analyses of the age (≤ 40, 41–54, 55–64, and ≥ 65) and sex subgroups revealed no significant sex-by-treatment or age-by-sex interactions; men and women of different ages within each treatment group had similar rates of remission, response, and absence of depressed mood symptoms.295
A pooled analysis of data from four sertraline-RCTs conducted in populations with panic disorder, however, reported better responses of female patients on some outcome measures (panic attack frequency, time spent worrying).301 No differences were apparent in quality of life measures.
Another pooled data analysis of four placebo-controlled duloxetine trials assessed safety and tolerability of duloxetine for the treatment of MDD in 560 men and 1,062 women.302 There were no clinically meaningful differences between men and women in safety and tolerability with duloxetine treatment. This analysis showed no significant differential sex effects for pulse, blood pressure or weight. Withdrawals due to adverse events were similar between men and women. The only significant difference was in the occurrence of nausea; the nausea rate among placebo-treated patients was significantly greater in females than in males (10.7% compared with 3.7%, P < 0.008).
In another pooled analysis of placebo-controlled trials of desvenlafaxine (n=2913) authors found a significantly higher risk of vomiting for women (OR, 3.36; 95% CI: 2.01–5.63) than for men (OR, 1.12; 95% CI, 0.47–2.63; P<0.03).303 For efficacy and other safety outcomes the study did not reveal any significant sex-treatment interactions.
One fair study randomized patients to bupropion (150–300 mg/d) or paroxetine (20–40 mg/d).253 Subgroup analysis revealed that a significant difference in anti-depressant related sexual dysfunction was detected in men but not in women. There were no significant drug differences between bupropion- and paroxetine-treated women in sexual function. However, paroxetine-treated men reported a worsening of sexual function while bupropion-treated men had no significant change in sexual function (Sex FX total, P<0.002).
A fair-rated meta-analysis304 included experimental and observational studies to assess differences in sexual dysfunction between men and women taking citalopram, fluoxetine, paroxetine, sertraline, or venlafaxine. All drugs caused significantly higher rates of orgasm dysfunction (citalopram OR 4.60, 95% CI: 3.01 to 7.02, P < 0.00001; fluoxetine: OR 6.00, 95% CI: 4.25 to 8.48; P < 0.00001; paroxetine: OR 5.60, 95% CI: 3.79 to 8.29; P < 0.00001; sertraline: OR 4.29, 95% CI: 3.01 to 6.12; P < 0.00001; venlafaxine: OR 7.60; 95% CI: 4.16 to 13.89; P < 0.00001) in men; for paroxetine and sertraline there was higher arousal dysfunction in women (paroxetine: OR 0.45, 95% CI 0.31 to 0.67; P < 0.0001; sertraline: OR 0.50, 95% CI 0.34 to 0.74; P < 0.0005).
In a study comparing fluvoxamine (50 mg/d) and paroxetine (20 mg/d), there was a significant difference in the decrease in hotflashes in menopausal women favoring paroxetine (−81.1 compared with −66.8, P<0.01).81 However, there were no statistically significant differences in depression symptoms.
B. Other Medications-Drug Interaction
The evidence for drug-drug interactions is limited. A 2004 study published in the Journal of the American Pharmacists Association reported that there was very little agreement in reporting clinical significance of drug-drug interactions.305 In fact, the authors found that only 2.2 percent of major drug interactions were listed in all sources reviewed.
Based on our review criteria, we did not identify any head-to-head trials specifically evaluating drug-drug interactions. We found one recent, fair quality population-based retrospective cohort study exploring the relationship between SSRI use and co-occuring tamoxifen use (a prodrug metabolized by the hepatic cytochrome P450 enzyme system) for breast cancer.306 The authors used data from 2430 women (median age 74 years in the year before starting tamoxifen) and included five SSRIs (citalopram, fluoxetine, fluvoxamine, paroxetine, sertraline) and venlafaxine in the analysis. They assessed death from breast cancer as a consequence of potential interaction between SSRIs and tamoxifen by cytochrome P450 inhibition. Risk of death from breast cancer in women receiving tamoxifen and paroxetine concurrently was significantly increased. The increased risk was directly related to the extent of co-prescribing. Absolute increases of 25 percent, 50 percent, and 75 percent in the proportion of time on tamoxifen that overlapped with use of paroxetine were associated with relative increases of 24 percent, 54 percent, and 91 percent in the risk of death from breast cancer, respectively (adjusted hazard ratios 1.24, 95% CI 1.08 to 1.42; 1.54, 95% CI 1.17 to 2.03, and 1.91, 95% CI 1.26 to 2.89, respectively). No such risk was found with citalopram, fluoxetine, fluvoxamine, sertraline, or venlafaxine.
Because only limited evidence supports drug interactions among the second-generation antidepressants, our review focuses on the potential for drug interactions. In addition to published literature cited previously, we reviewed dossiers submitted by pharmaceutical companies, FDA approved labeling, and interactions reported by major reference sources. Information compiled in this search does not follow a systematic process but is provided as a summary of the evidence for drug interactions. Appendix D summarizes second-generation antidepressant pharmacokinetic properties known to be related to drug interactions. Tables in Appendix D report evidence provided in the product labeling (package insert). Some interactions are inferred based on reports of enzyme induction or inhibition. Clinical significance of the interactions are referenced as contraindicated, requires monitoring, or no significant interaction.
We found no studies directly comparing the efficacy, effectiveness, and tolerability of second-generation antidepressants between depressed patients with comorbidities and the general population. Therefore, we only describe studies conducting subgroup analyses or studies using subgroups as the primary study population. In addition, we do not present findings under subheadings of drug classes for each comorbid condition.
1. Chronic conditions combined
SSRIs compared with placebo
A good meta-analysis using data from six placebo-controlled RCTs on 1299 patients with long-term SSRI-therapy (citalopram, paroxetine, sertraline) for the treatment of depression conducted a subgroup analysis of RCTs in patients with major chronic health conditions (myocardial infarction, stroke) and alcohol dependence.307 Authors found that with respect to response, overall SSRIs were superior to placebo at 6 to 8 months (OR 1.66, 95 CI 1.12 to 2.48), but not among patients with comorbidities (OR 1.32, 95% CI 0.84 to 2.06). Also, participants without comorbidities had a significantly higher remission rate if treated with SSRIs as compared to those in the placebo group (OR 2.06, 95% CI 1.41 to 3.01); no such statistically significant treatment effect was found in participants with comorbidities (OR 0.87, 95% CI 0.44 to 1.72). Across the trials, the mean dropout rate was 48 percent (range 27%-77%) and authors rated the quality of the included trials as moderate.
2. Alcohol/substance abuse
Fluoxetine compared with placebo
Five randomized placebo-controlled trials assessed the efficacy and tolerability of fluoxetine for the treatment of depression with co-occurring alcoholism308–311 or co-occurring substance use disorders.312–314
One fair study of 51 depressed alcoholics assessed the efficacy of fluoxetine (20–40 mg/d) in a 12-week, placebo-controlled, acute-phase trial and a subsequent 1-year follow-up period with a naturalistic treatment by physicians unrelated to this study (N=31).308–310 Outcome measures included changes on HAM-D and BDI and in alcohol consumption. Results of the acute phase trial showed significantly greater improvements of depressive symptoms for fluoxetine-treated patients (P<0.05) on HAM-D but not on BDI. During the 1-year open-label follow-up, HAM-D scores remained significantly lower for the fluoxetine group than for the placebo group. However, no additional improvement during the follow-up treatment was reported. A subgroup analysis showed that depressed alcoholics who were cocaine abusers (N=17) had a significantly worse outcome than depressed alcoholics who were not (N=34). Cocaine abusers showed significantly worse outcomes on both the HAM-D (P=0.17) and the BDI (P=0.001).
A fair, small RCT assessed the efficacy and tolerability of fluoxetine treatment (20–60 mg/d) compared to placebo for the treatment of major depression in 44 methadone-maintained opioid addicts.312 Study duration was 3 months; loss to follow-up was 15.9 percent. Both groups had significantly decreased scores on BDI and HADRS (z = 2.37; P=0.01). Efficacy did not differ significantly between placebo and fluoxetine treatment. However, the sample size was small and the study is likely to be underpowered (no power calculations were reported).
A poor quality study investigated the efficacy of fluoxetine (40 mg/d) in 68 cocaine-dependent patients with MDD.313 The trial was rated poor for efficacy due to its high attrition rate (53%), but we included it here because of the dearth of evidence on this topic. Results showed no difference in efficacy between fluoxetine and placebo at the end of this 12-week study.
One fair 16-week RCT assessed the efficacy and tolerability of fluoxetine (20 mg/d) plus cognitive behavior therapy compared with placebo plus cognitive behavior therapy in 126 adolescents (mean age 17.2 years) with MDD and comorbid substance abuse disorder and conduct disorder.314 Decreases in Childhood Depression Rating Scale-Revised (CDRS-S) scores were greater in fluoxetine- than placebo-treated patients (−22.5 compared with −16.6) Fluoxetine-treated patients showed a greater CGI-I response than placebo patients, but the difference was not statistically significant (76.3% compared with 66.7%, RR = 1.14). There were no differences between groups in substance abuse disorder, conduct disorder or urine drug screen. In addition, there were no differences between groups in the incidence of adverse events.
A small, fair-rated 12-week RCT of 50 patients compared the efficacy of fluoxetine (20mg/d) versus placebo for the treatment of depressive symptoms and drinking behavior in adolescents (15–20 years of age) with comorbid MDD and an alcohol use disorder.311 All study participants also received sessions of cognitive behavioral therapy and motivation enhancement therapy. While participants in both arms experienced improvements in depressive symptoms and drinking-related outcomes, no significant differences in depressive symptoms or drinking behavior between the treatment groups were found.
Nefazodone compared with placebo
One randomized trial compared nefazodone and placebo in the treatment of depressed patients with depression and comorbid alcohol dependence over a 10-week period.315 HAM-D scores at endpoint showed no significant difference between treatment groups in depressive symptoms (P=0.51). Nefazodone-treated subjects averaged 0.8 fewer heavy drinking days per week than placebo-treated subjects (P=0.01). More nefazodone-treated patients were abstinent during treatment; however, the difference did not reach statistical significance (P=0.17).
Paroxetine compared with placebo
A fair study randomized 42 subjects with social anxiety disorder and a co-occurring alcohol use disorder to paroxetine (10–60 mg/d) or placebo for 16 weeks.316 Decreases in total LSAS scores were significantly greater for paroxetine- compared to placebo-treated patients (53% compared with 32%, P=0.02). A higher percentage of paroxetine-treated patients were CGI responders (defined as improvement score of 1 or 2) compared to placebo-treated patients (55% compared with 27%). The mean reductions in Social Phobia Inventory (SPIN) results were greater in the paroxetine group but did not reach statistical significance (46% compared with 31%, P=0.15). Three specific adverse events occurred significantly more frequently in paroxetine-treated patients: tremor (45% compared with 14%, P=0.03), myoclonus (35% compared with 5%, P=0.01) and anorgasmia/delayed ejaculation (55% compared with 18%, P=0.01).
Sertraline compared with placebo
Three fair RCTs compared sertraline and placebo in the treatment of patients with depression and co-occurring alcohol dependence.317–319
A 24-week study compared sertraline (50–150 mg/d) with placebo in recently detoxified alcohol-dependent patients with current depressive symptoms.317 Response (≥ 50% decrease in MADRS score) was slightly higher in sertraline- than placebo-treated patients (44% compared with 39%). Both groups experienced significant improvements in HAM-D and MADRS scores during the study, but the two groups did not differ significantly. Relapse rates were higher in sertraline- than placebo-treated patients (31.8% compared with 23.1%) but the difference was not statistically significant (P=0.37). Adverse event rates were similar for both treatment groups. The overall attrition rate was greater than 40 percent; however, there was not a significant difference in withdrawal between groups (sertraline, 45% compared with placebo, 44%).
A 12-week trial showed similar results.318 In this fair study, 82 currently depressed, actively drinking alcohol-dependent subjects were randomized to sertraline (50–200 mg/d) or placebo. There was no significant difference between groups in depression symptoms. However, in women, treatment with sertraline was associated with less depression at the end of treatment than those receiving placebo based on HAM-D scores (P=0.04) and BDI scores (P=0.005). There was no treatment group difference for men. There was no difference between groups in time to first heavy drinking day (P=0.661) or days abstinent or heavy drinking days per week. Sertraline-treated subjects had fewer drinks per drinking day compared to placebo-treated subjects; the difference was significant (P=0.27). Less drinking during the study was associated with improved depression outcomes. Serious adverse events occurred in four subjects: three treated with sertraline and one treated with placebo. Loss to follow-up was twice as high in the placebo group (33%) compared to the sertraline group (16%); however, details were not reported on withdrawals due to tolerability or lack of efficacy.
The third study was structured differently but produced similar results.319 This study randomized 328 patients with co-occurring MDD and alcohol dependence to sertraline (50–200 mg/d) or placebo for 10 weeks. After the run-in period, two groups of patients were randomized separately based on HAM-D scores: Group A scores were ≥ 17 while Group B scores were ≤ 16. Mean reduction in HAM-D scores did not differ significantly between all sertraline-treated (−10.8) and placebo-treated (−9.6) patients (P=0.14). There were significant differences in HAM-D response rates by group stratification. In Group A, sertraline led to significantly higher response rate than placebo (64% compared with 47%, P=0.022). However, in Group B, sertraline patients had a significantly lower response rate than placebo patients (58% compared with 77%, P=0.018). There were no significant differences between medication groups in the reduction in BDI score from baseline to endpoint nor within Group A or Group B. No significant differences were detected between medication groups in drinking measures. Overall, the incidence of adverse events was similar between medication groups; however, significantly more sertraline-treated patients discontinued due to adverse events than placebo-treated patients (P<0.05).
3. Alzheimer’s disease/dementia
The first,320 a fair 12-week trial, demonstrated that sertraline was statistically significantly superior to placebo as measured by both the Cornell Score for Depression in Dementia (CSDD) and the HDRS (P<0.01). More patients treated with sertraline responded to treatment (full responders, 38%; partial responders, 46%) than did patients treated with placebo (full responders, 20%; partial responders, 15%) (P<0.007).
A second fair 12-week trial which randomized 133 patients with mild-to-moderate Alzheimer’s disease and depression to either sertraline (100mg/d) or placebo did not replicate the above findings. Mood was assessed by the modified Alzheimer Disease Cooperative Study-Clinical Global Impression of Change index and the CSDD. At the end of week 12, CSDD scores and remission rates did not differ between sertraline and placebo (OR 2.06, 95% CI 0.84 to 5.04, P < 0.11) with a high percentage of patients in both groups experiencing clinical improvements. Treatment with sertraline, however, was associated with more adverse events, specifically gastrointestinal adverse events than with placebo. Serious adverse events occurred in 20 percent of patients in the sertraline group compared with 11 percent in the placebo group.321
Our searches yielded only one trial that evaluated the efficacy of an antidepressants in depressed patients with comorbid arthritis.322 This study is a subgroup analysis of a larger placebo-controlled trial in elderly patients randomized to duloxetine (60 mg/d) or placebo.323 The subgroup analysis analyzed 233 subjects with MDD and co-occurring arthritis, diabetes and/or vascular disease; 55 percent of patients had diabetes. There were no statistically significant treatment-by-comorbidity interactions for any comorbidity (P < 0.266) in HAM-D, GDS, or SF-36 scores or in response or remission rates. Results must be interpreted with caution as this was the only study addressing this topic.
Fluoxetine compared with placebo
We detected only one trial that studied the efficacy of fluoxetine in cancer patients;324 however, this placebo-controlled trial failed to meet our inclusion criteria because the duration of the study was less than 6 weeks. We mention it here because it was the only trial on this topic. This 5-week trial studied the efficacy of fluoxetine in 91 cancer patients with depression or adjustment disorder. The majority of the patients were female; 13 percent in the fluoxetine group and 5 percent in the placebo group had metastatic disease. Outcome measures included quality of life. Loss to follow-up was 24.2 percent. Efficacy according to the main, observer-rated outcome measures (HADS, MADRS, HAS) did not differ significantly between the active drug and placebo groups. Improvements were generally greater in the fluoxetine group but statistically significant only for the SCL90-R (33% compared with 15%; P=0.04), which measures global psychological adjustment. No statistically significant difference in quality of life was reported. However, study duration was short and a substantially greater percentage of patients in the fluoxetine group had a more advanced stage of cancer at baseline. Fluoxetine-treated patients had a significantly greater drop-out rate than placebo-treated patients (33% compared with 15%; P=0.04).
Paroxetine compared with placebo
A 6-week randomized trial compared paroxetine (20 mg/d) and placebo in depressed breast cancer patients who were receiving at least four cycles of chemotherapy to evaluate whether the use of an antidepressant can alleviate symptoms of depression and reduce fatigue.325 Although this study was rated poor because of lack of ITT analysis and inadequate description of study duration, we included it because it was the only study conducted in cancer patients that satisfied our inclusion criteria. Paroxetine was more effective in reducing depression during chemotherapy, as measured by the Center for Epidemiological Studies of Depression (CES-D) (P=0.006). No differences between treatment groups were apparent with respect to fatigue.
Our searches yielded two trials that evaluated the efficacy of an antidepressants in depressed patients with comorbid diabetes.322, 326 One fair-rated study randomized 89 depressed, low-income Hispanics and African Americans with diabetes to sertraline (50–100 mg/day) or placebo for 6 months.326 HAM-D scores decreased significantly in both groups but there was no significant difference between sertraline-treated and placebo-treated patients. Quality of life measures improved significantly in both groups, but no difference was found between groups.
The details of the second study322 are described above (in the arthritis section). Only 15 percent of patients had comorbid diabetes mellitus. There were no statistically significant treatment-by-comorbidity interactions for any comorbidity (P < 0.266) in HAM-D, GDS, or SF-36 scores or in response or remission rates. Results must be interpreted with caution based on the small percentage of patients in this study who had comorbid diabetes in this study.
A fair placebo-controlled trial lasting 8 weeks determined the efficacy of fluoxetine (dosage range not reported) in 120 depressed patients with HIV and AIDS.327 The majority of patients were male (97.3%) and white (65%). Loss to follow-up was 27.5 percent. The main outcome measures were response to treatment defined as a 50 percent improvement on the HAM-D scale, a score lower than 8, and a CGI score of 1 or 2. According to these criteria, the rate of response did not differ significantly between treatment groups (fluoxetine 57%, placebo 41%). Using the HAM-D scale alone as a criterion, the investigators reported a significantly greater response rate for fluoxetine-treated patients (79% compared with 57%; P=0.03). The treatment groups did not differ significantly in adverse events.
The second trial (described above for ethnicity) evaluated the efficacy and tolerability of fluoxetine (20–80 mg/day) and placebo in depressed patients with comorbid HIV/AIDS. This study was rated poor because it had no ITT analysis; however, we included it here because of the very limited evidence on this topic.298 Response rates among subjects who completed the study were higher in the fluoxetine group than in the placebo group; however, the differences were not significant.
8. Multiple sclerosis
We detected only one study assessing the efficacy and tolerability of antidepressants for depression with comorbid multiple sclerosis (MS).328 Forty-two MS patients diagnosed with MDD and/or dysthymia were randomized to paroxetine (10–40 mg/d) or placebo for 12 weeks. Although more paroxetine-treated patients achieved at least a 50 percent reduction in HAM-D scores (57%) compared to placebo-treated patients (40%), the difference was not statistically significant (P=0.354). Paroxetine- and placebo-treated patients showed improvement in secondary measures (CES-D, MFIS, SF-36), but there were no significant differences between treatment groups. Paroxetine patients reported higher rates of nausea, headache, dry mouth and sexual dysfunction.
9. Somatizing depression
A retrospective evaluation of 89 patients from two trials comparing fluoxetine (20–80 mg/d) to paroxetine (20–50 mg/d) determined whether depressed, somatizing patients with a gastrointestinal (GI) component have a higher degree of GI side effects than nonsomatizing depressed participants.329 Participants with baseline complaints of nausea, upset stomach, GI somatic symptoms, or weight loss were not statistically more likely to develop additional GI side effects than those without such complaints at the start of the trials.
10. Vascular disease (cardiovascular, cerebrovascular or peripheral vascular)
We identified eight placebo-controlled trials,322, 330–336 one pooled-data analysis,337 and one systematic review338 that addressed depression and co-occurring vascular disease of some type (chronic heart failure, coronary artery disease, post-myocardial infarction, stroke, and vascular disease). The majority of the trials evaluated a different drug (citalopram, duloxetine, fluoxetine, mirtazapine, sertraline, and SSRIs as a class) with the exception of sertraline—two studies compared sertraline and placebo. Therefore, results are presented here by comorbidity rather than by drug comparison.
Chronic heart failure
We detected one study evaluating comorbid chronic heart failure in depressed patients.339 However, this study did not meet our inclusion criteria due to its small sample size. We discuss it here because of the paucity of evidence on this topic. In this study, 28 patients with symptomatic congestive heart failure and MDD were randomized to paroxetine CR (25 mg/d) or placebo for 12 weeks. Paroxetine resulted in significantly more remission of depression (BDI < 10) than placebo (69% compared with 23%, P=0.018). Paroxetine was superior to placebo in quality of life changes based on overall SF-36 scores (P<0.05). Reductions in SF-36 scores did not correlate with improvements in physical quality of life measures (P>0.10). There were no differences in adverse events. Valid conclusions cannot be drawn, however, because of the small sample size in this study.
Coronary artery disease
One fair 12-week Canadian study assessed the efficacy and tolerability of citalopram (20–40 mg/d) and placebo in reducing depressive symptoms in patients with co-occurring coronary artery disease (CAD).330 Improvements in depressive symptoms were greater for citalopram than placebo. Mean HAM-D24 scores at endpoint showed significantly greater improvement in citalopram-treated patients compared to placebo-treated patients (14.9 compared with 11.6, P=0.005); between group difference was 3.33 (95% CI 0.80 to 5.85). Citalopram-treated patients also demonstrated significantly greater decrease in mean BDI-II scores at endpoint (P<0.05); between group difference was 3.61 (95% CI 0.58 to 6.64). Incidences of six adverse events were significantly greater in citalopram-treated patients: dizziness (48.6% compared with 30.3%, P=0.002), diarrhea (49.3% compared with 23.9%, P<0.001), somnolence (43.7% compared with 25.4%, P=0.001), sweating (39.4% compared with 23.9%, P=0.005), palpitations (25.4% compared with 14.8%, P=0.003), and decreased libido or sexual difficulties (21.1% compared with 7.0%, P=0.001). The citalopram group had a lower overall withdrawal rate (13% compared with 30%, P=NR); however, withdrawals due to adverse events were similar between treatment groups.
Three placebo-controlled trials and one systematic review evaluating second-generation antidepressants in the treatment of comorbid post-myocardial infarction.(post-MI). A fair quality systematic review sponsored by AHRQ examined the role of depression in post-MI.338 One section of this review addressed SSRI treatment for post-MI depression and included 11 studies. The authors concluded that SSRIs improve depression in post-MI patients and some surrogate markers of cardiac risk. However, the authors also found that none of the studies was powered to show whether treatment improves survival. The authors did not address the tolerability of SSRIs in their review.
A 24-week trial randomized 369 patients with MDD and acute MI or unstable angina to sertraline (50–200 mg/d) or placebo.335 Sertraline was associated with a significantly greater percent of CGI-I responders compared to placebo (67% compared with 53%, P = 0.01). However, there was not a significant difference between groups in mean change in HAM-D score (P = 0.14). The incidence of severe cardiovascular adverse events was lower in sertraline patients (15% compared with 22%), but the difference was not significant. Both nausea and diarrhea were significantly more common in sertraline patients (P=NR).
The second, a good quality trial randomized 54 depressed patients after a first MI to fluoxetine (20–60 mg/d) or placebo for 25 weeks (9 weeks of acute treatment and an additional 16 week continuation phase).332, 340 Significantly more sertraline-treated patients were HAM-D responders compared to placebo-treated patients after 25 weeks (48% compared with 26%, P = 0.05). In addition, sertraline patients showed a greater mean decrease in SCL-90 hostility scores (−2.44 compared with −0.07, P=0.02). Percent of HAM-D remitters and mean decreases in HAM-D score also favored sertraline; however, differences did not reach statistical significance. One sertraline- and six placebo-treated patients were rehospitalized for a cardiac event during the study (P=0.13).
The third study randomized 91 patients to mirtazapine (30–45 mg/d) or placebo for 8 weeks of acute treatment (and a 16-week continuation phase).334 After 8 weeks of treatment, mirtazapine was superior to placebo based on BDI and CGI scales but not HAM-D. The difference between treatment groups in mean decrease in HAM-D score was not significant at 8 weeks (standardized effect size [SES] 1.30 compared with 0.96). Based on change in HAM-D score at 8 weeks, more mirtazapine-treated patients were responders (57% compared with 40%), but the difference was not significant (P=0.18). Mirtazapine-treated patients showed a significantly greater decrease in BDI score at 8 weeks (−4.6 compared with −1.72, P=0.02). Decrease in CGI score was greater in mirtazapine-treated patients but the difference was not statistically significant (P=0.06). The differences between groups in decrease in HAM-D scores and BDI scores over 24 weeks was not statistically significant (P=0.36 and P=0.07). The difference in CGI-scores over 24 weeks favored mirtazapine; the difference was significant (P=0.05). Mirtazapine patients experienced significantly more fatigue (P=0.02) and changes in appetite (P=0.02) over 24 weeks.
One fair 6-week randomized trial evaluated the efficacy of citalopram (10–40 mg/d) and placebo in the treatment of 66 patients with poststroke depression.331 Citalopram was associated with significantly greater improvements in depression compared to placebo on the HAM-D; mean (SD) improvements for citalopram compared with placebo were 8.0 (6.0) and 7.2 (5.8), respectively.
A fair 6-week trial of 150 patients assessed the efficacy and tolerability of fluoxetine (20–40 mg/d) compared with placebo and with a Chinese herbal formula in the treatment of post-stroke depression.333 The fluoxetine and placebo groups consisted of 90 patients, all of which had a recent single ischemic or hemorrhagic stroke. Significantly higher clinical response rates were observed in the fluoxetine compared with the placebo group (60% versus 21.4%, χ2= 15.9, P < 0.01). No serious side effects were reported.
A fair 26-week trial evaluated the efficacy and tolerability of sertraline (60–100 mg/d) compared with placebo in the treatment of minor depression and less severe depression in 123 stroke patients.336 Sertraline and placebo patients improved substantially but did not differ significantly in HAM-D response rates (76% compared with 78%) or in MADRS remission rates (81% compared with 87%). However, at week 26, sertraline was associated with greater improvements in quality of life than placebo (effect size not reported, P<0.05). Sertraline-treated patients experienced higher rates of three adverse events compared to placebo-treated patients: dry mouth (23.6% compared with 7.4%, P<0.05), diarrhea (23.6% compared with 9.3%, P<0.05), and emotional indifference (9.1% compared with 0%, P<0.05).
Peripheral vascular disease
We detected two trials addressing the efficacy of depressed patients with comorbid vascular disease.322, 337 One trial that evaluated the efficacy of duloxetine (60 mg/d) and placebo in elderly patients.322 The details of this study are described above (in the KQ3 arthritis section). In this study, 75 percent of the patients had comorbid vascular disease. There were no statistically significant treatment-by-comorbidity interactions for any comorbidity (P=0.266) in HAM-D, GDS, or SF-36 scores or in response or remission rates. Results must be interpreted with caution based on the small percentage of patients in this study who had comorbid diabetes in this study.
The second study, a fair, retrospective analysis of pooled data of two RCTs determined the safety and efficacy of sertraline (50–150 mg/d) in elderly patients with comorbid vascular disease.337 Vascular comorbidity was not associated with an increase of severity of adverse events or premature discontinuation. However, these findings were not based on an unbiased literature search and the validity must be viewed cautiously.
D. Summary of the Evidence
We found no study that directly compared efficacy and safety of treatments in an elderly population compared to a younger population. A fair pooled data analysis did not find significant associations between age and outcomes or age and treatment.294 However, findings suggested that older women had a poorer response to SSRIs than younger women.294, 295
Eight studies provide fair to good indirect evidence that efficacy and tolerability for patients older than 60 years and those younger do not differ.46, 55, 65, 68, 77, 79, 89, 92, 111, 112 Results of these studies, all conducted in patients with MDD or dysthymia, are generally consistent with results of trials conducted in younger populations. Only one small study reported a higher efficacy of paroxetine than fluoxetine in patients older than 60 years.71 However, this trial was small and the results are inconsistent with better evidence. Another small study, rated poor for efficacy outcomes, reported a significantly higher loss to follow-up because of adverse events in venlafaxine-treated, frail elderly patients than in sertraline-treated participants.293
For MDD, placebo-controlled evidence supports the efficacy of fluoxetine341, 342 and sertraline.156 Existing evidence does not support the efficacy of other second-generation antidepressants. Additional evidence suggests that sertraline may not be as efficacious as reported in previous reports. Based on one systematic review of published and unpublished studies comparing second-generation antidepressants to placebo, only fluoxetine was shown to be safe and effective in the treatment of MDD in children and adolescents.146 This review reported an increased risk of suicidal thoughts and behavior for citalopram, paroxetine, sertraline, and venlafaxine, but not for fluoxetine. Two other systematic reviews of confirmed these results finding only fluoxetine had a favorable risk-benefit profile.147, 148
Fair evidence from a pooled data study on paroxetine299 and a single RCT on fluoxetine298 suggest that response rates, loss to follow-up, and response to placebo treatment might differ between groups of different ethnic background. Hispanics tend to have lower response rates than Blacks and Whites. However, two pooled data analyses (of the same seven placebo-controlled duloxetine trials) found no significant differences between Caucasians and Hispanics296 or between Caucasians and African Americans.297 Altogether, the evidence is inconclusive to determine whether second-generation antidepressants differ between patients with diverse ethnic backgrounds.
Two pooled-data analyses did not find significant associations between sex and efficacy outcomes in patients treated for MDD.294, 295, 303 A pooled analysis of data from four sertraline-RCTs conducted in populations with panic disorder reported better responses of female than male patients on some outcome measures.301
A fair trial comparing bupropion and paroxetine showed a significant difference in anti-depressant related sexual dysfunction in men but not in women. Paroxetine-treated men reported a worsening of sexual function while bupropion-treated men had no significant change in sexual function. A meta-analysis of RCTs found significant gender-related adverse events of antidepressants. Citalopram, fluoxetine, paroxetine, sertraline, and venlafaxine caused higher rates of desire and orgasm dysfunction in men and higher arousal dysfunction in women.304 A pooled data analysis indicated higher rates of vomiting in women than in men treated with desvenlafaxine.303
4. Concomitant medications
A fair retrospective cohort study found evidence of increased breast cancer mortality in women treated with tamoxifen for breast cancer and concurrent use of paroxetine. No evidence of increased risk was found with concurrent use of fluoxetine, sertraline, citalopram, fluvoxamine, or venlafaxine.306 Evidence is insufficient to determine the influence of concomitant medications on the effectiveness or harms of SSRIs, SNRIs, or other second-generation antidepressants.
We found no prospective study directly comparing the efficacy, effectiveness and tolerability of SSRIs, SNRIs, and other second-generation antidepressants in a population with a specific comorbid condition to a population without that same condition. A meta-analysis with a subgroup analysis yielded good evidence that overall, SSRIs were superior to placebo at 6 to 8 months for patients without comorbidities compared with patients with comorbidities (analyzed as a combined category).307 However, we could not identify further studies analyzing outcomes after a follow-up of similar duration.
The majority of studies we identified are limited to depressive disorders in patients with a variety of disorders. Overall, evidence of treatment efficacy across various disease conditions and second-generation depressants is mixed.
For some comorbidities such as post-myocardial infarction335, 338 or coronary artery disease 330 the evidence indicates a general efficacy of some second-generation-antidepressants for the treatment of depression.
For other conditions, however, such as methadone-maintained opioid addiction, cocaine abuse, HIV, multiple sclerosis, arthritis, diabetes, or cancer,312, 313, 322, 324, 326–328 comorbid alcohol use disorder in depressed adolescents311 or substance abuse in adolescents with comorbid conduct disorder,314 second-generation antidepressants were unable to achieve response or remission rates that were statistically significantly better than placebo.
For some comorbid conditions the evidence was inconclusive. For depression with comorbid alcoholism, evidence of treatment efficacy of a SSRI compared with placebo,308–310 lack of such an effect,317, 319 or differential effect only for women in the treatment group318 were identified. Also, treatment efficacy for post-stroke depression was not uniform across studies, with two trials showing second-generation antidepressant superior to placebo331, 333 yet one trial336 lacking a significant treatment effect. Inconclusive also the findings of trials with second-generation-antidepressants for Alzheimer’s depression, with one study showing a treatment effect,320 a second trial, however, lacking such an effect but with more adverse events in the treatment group.321
Oregon Health & Science University, Portland (OR)
Gartlehner G, Hansen RA, Reichenpfader U, et al. Drug Class Review: Second-Generation Antidepressants: Final Update 5 Report [Internet]. Portland (OR): Oregon Health & Science University; 2011 Mar. Results.