We found 7 systematic reviews
All the reviews found that antidepressant drugs significantly improved depressive symptoms compared with placebo.
The first review (search date 1995, 49 RCTs in people aged 18–70 years with mild to moderate
depressive disorders) included 5 RCTs in people admitted to hospital (probably with severe depressive disorders), 40 RCTs in a setting outside hospital, one RCT in both settings, and three RCTs that did not specify the setting.
All RCTs identified by the review were of at least 4 weeks' duration and included three-way comparisons, including two antidepressant drugs (tricyclic antidepressants [TCAs], SSRIs, or monoamine oxidase inhibitors [MAOIs]) and placebo. The review only included RCTs that measured improvement in depressive symptoms using validated scales such as the Hamilton Depression Rating Scale (HAM-D) and Montgomery–Asberg Depression Rating Scale. It found that, on average, 69% of people taking placebo had worse outcomes over a mean of 6 weeks than the average person taking antidepressant drugs (mean effect size
0.5 for change in score with antidepressant drugs v
The second review (search date 2000, 35 RCTs, none included in the first review, 2013 people aged 18 years or older with all grades of depression, some with a physical illness) compared low-dose (75–100 mg/day) TCAs (amitriptyline, clomipramine, doxepin, dosulepin [dothiepin], imipramine, lofepramine, trimipramine) versus placebo.
It found that low-dose TCAs significantly increased the proportion of people who responded at 4 weeks and at 3 to 12 months compared with placebo (response defined as 50% or greater reduction in symptoms measured on a validated scale; at 4 weeks: 274/603 [45%] with TCAs v
159/557 [28%] with placebo; RR 1.65, 95% CI 1.36 to 2.00; at 3–12 months: 40/76 [53%] with TCAs v
18/77 [23%] with placebo; RR 2.14, 95% CI 1.41 to 3.26; results not intention to treat). The significant difference in response rates between low-dose TCAs and placebo was not maintained when an intention-to-treat analysis based on the worst case scenario was performed.
The third review (search date 1998, 150 RCTs, 16,000 people or greater with major depression) compared newer antidepressants (SSRIs [43 RCTs], MAOIs, or venlafaxine) versus placebo for at least 6 weeks.
Response was defined as a 50% reduction in depression rating scale score or a Clinical Global Impression Scale (CGI) score of 1 (very much improved) or 2 (much improved). The review found that newer antidepressants significantly increased the proportion of people who responded compared with placebo (51% with newer antidepressants v
31% with placebo; RR 1.6, 95% CI 1.5 to 1.7). The third review
also performed a separate analysis of results for people in primary care.
It found that results remained significant; the average response rate was 63% with newer agents, 35% with placebo, and 60% with TCAs (RR for SSRIs v
placebo 1.6, 95% CI 1.2 to 2.1).
The fourth review (search date 1997, 15 RCTs, none included in the other reviews, some comparing two antidepressants v
placebo, 1871 people aged 18 years or older) compared antidepressant and other drugs (TCAs [5 RCTs], SSRIs [4 RCTs], MAOIs [3 RCTs], other [2 RCTs]) versus placebo in people with dysthymia (chronic mild depressive disorders).
It found that antidepressant or other drugs significantly increased the proportion of people who responded to treatment over 4 to 12 weeks compared with placebo (response defined as a 50% reduction in HAM-D score or scoring 1 or 2 on item 2 of the CGI score; RR 1.9, 95% CI 1.6 to 2.3; NNT 4, 95% CI 3 to 5).
The fifth review found that SSRIs plus pindolol (7.5–15.0 mg/day) significantly increased the proportion of people with early clinical response compared with SSRIs plus placebo (search date 2001, 5 RCTs, 405 people aged >18 years with depressive illness; proportion with early clinical response: 66/187 [35%] people with SSRIs plus pindolol v
32/187 [17%] people with SSRIs plus placebo; OR 2.8, 95% CI 1.4 to 5.7; NNT 6, 95% CI 4 to 20).
The review found no significant difference between SSRIs plus pindolol and SSRIs plus placebo in late clinical response (7 RCTs, 187 people aged >18 years with depressive illness; proportion with late clinical response: 142/208 [68%] people with SSRIs plus pindolol v
124/206 [60%] people with SSRIs plus placebo; OR 1.4, 95% CI 0.8 to 2.7).
The sixth review (search date 2004, 15 RCTs, 2753 people aged 18 years or older in primary care for depression) compared TCAs versus placebo (10 RCTs), SSRIs versus placebo (3 RCTs), and TCAs and SSRIs versus placebo (2 RCTs).
The review found that both TCAs and SSRIs were significantly more effective at treating depression compared with placebo (TCA v
placebo: SMD for depression scores –0.42, 95% CI –0.55 to –0.30; RR for improvement 1.26, 95% CI 1.12 to 1.42; 323/535 [60%] with TCAs v
216/460 [47%] with placebo; SSRI v
placebo: RR for improvement 1.37, 95% CI 1.21 to 1.55; 310/552 [56%] with SSRIs v
231/562 [41%] with placebo). There were several definitions used to describe "improvement": 4 definitions used >50% reduction in the Montgomery–Asberg Depression Rating Scale (MADRS), >50% reduction in HAM-D, >7 on the HAM-D scale, and 4 points or greater on HAM-D; and three definitions used global evaluation of improvement.
The seventh review (search date 2004) was a high-quality systematic review undertaken as part of a guideline development process.
The guideline, initially published in 2004, was amended in 2007 and further updated in 2009 at which time some, but not all, of the included analyses were updated (to search date 2009).
The review found that TCAs significantly improved depressive symptoms compared with placebo whether measured by dichotomous outcomes or by continuous outcomes (dichotomous: not achieving response [50% reduction in depression symptoms as measured by standardised rating scale], 34 RCTs, 4717 people; RR 0.70, 95% CI 0.66 to 0.75; continuous: mean depression scores at end point, 22 RCTs, 2445 people; SMD –0.48, 95% CI –0.37 to –0.59). The review found that SSRIs significantly improved depressive symptoms compared with placebo, whether measured by dichotomous outcomes or continuous outcomes (dichotomous: not achieving at least 50% reduction in depression scores measured by the Hamilton Depression Rating Scale, 17 RCTs, 3143 people; RR 0.73, 95% CI 0.69 to 0.78; continuous: mean end point scores measured by the Hamilton Depression Rating Scale, 16 RCTs, 2223 people; SMD –0.34, 95% CI –0.47 to –0.22). The results were similar when only RCTs lasting 8 weeks or longer were assessed, with SSRIs significantly more effective than placebo (not achieving at least 50% reduction in depression scores measured by the Hamilton Depression Rating Scale, 8 RCTs, 1764 people; RR 0.72, 95% CI 0.66 to 0.79). The results were similar when analysed by severity of illness: moderate depression (3 RCTs, 729 people; RR 0.75, 95% CI 0.65 to 0.87; 2 RCTs, 386 people; SMD –0.28, 95% CI –0.48 to –0.08); severe depression (5 RCTs, 619 people; RR 0.63, 95% CI 0.54 to 0.73; 4 RCTs, 344 people; SMD –0.61, 95% CI –0.83 to –0.40); and very severe depression (6 RCTs, 866 people; RR 0.72, 95% CI 0.65 to 0.8; 5 RCTs, 726 people; SMD –0.39, 95% CI –0.54 to –0.24). The review found insufficient evidence to determine whether there was a clinically significant difference between SSRIs and placebo in increasing the likelihood of achieving remission as measured by the HRSD (3 RCTs, 468 people; RR 0.80, 95% CI 0.61 to 1.06).
The analysis of SSRIs as a class versus placebo did not include escitalopram, the most recently marketed SSRI.
The update of the guideline considered escitalopram separately versus placebo, but results were not substantially qualitatively different from other SSRIs.
The review also reported on moclobemide and phenelzine. It reported that none of the included studies described participants as having depression with atypical features. The review found that, compared with placebo, moclobemide significantly reduced symptoms of depression by the end of treatment as measured by the HRSD (3 RCTs, 490 people; SMD –0.6, 95% CI –1.13 to –0.07) and significantly increased the likelihood of achieving at least a 50% reduction in symptoms as measured by the HRSD (3 RCTs, 606 people; RR 0.7, 95% CI 0.5 to 0.99). It found no placebo-controlled RCTs of phenelzine.
In older adults:
The first systematic review
compared antidepressant drugs versus placebo. The review found that TCAs, SSRIs, or MAOIs significantly reduced the proportion of people who failed to recover over 4 to 7 weeks compared with placebo (search date 2000, 17 RCTs, 1326 people aged >55 years, with mild to moderate or severe depression; numbers failing to recover: 125/245 [51%] with TCAs v
167/223 [75%] with placebo; RR 0.68, 95% CI 0.59 to 0.78; NNT 4, 95% CI 4 to 5; 261/365 [72%] with SSRIs v
310/372 [83%] with placebo; RR 0.86, 95% CI 0.79 to 0.93; NNT 9, 95% CI 9 to 10; 34/58 [59%] with MAOIs v
57/63 [90%] with placebo; RR 0.64, 95% CI 0.50 to 0.81; NNT 4, 95% CI 3 to 4).
The second systematic review (search date 2007, 22 RCTs) compared antidepressant drugs versus placebo in depressed people in later life.
It did not pool data, and did not report methodological or numerical data on included RCTs, or report on the antidepressant used. It reported that "some of the trials have shown efficacy of antidepressants against placebo (18 RCTs) whilst others have not (4 RCTs)" (absolute data and statistical analysis not reported).
The first subsequent RCT (264 people, 12 weeks' duration) compared escitalopram versus placebo in depressed people aged 60 years or older.
The primary end point was the mean change from baseline to week 12 in the MADRS total score. The RCT found no significant difference between escitalopram and placebo in response measured by MADRS (mean difference –1.34, 95% CI –3.84 to +1.15; P = 0.29).
The second subsequent RCT (525 people, 10 weeks' duration) compared two different low doses of controlled released (CR) paroxetine (CR 25 mg, 177 people; CR 12.5 mg, 168 people) versus placebo (180 people) in depressed people aged 60 years or older.
It found that, compared with placebo, a significantly higher proportion of people achieved remission (HAM-D score 7 or less at end point) with paroxetine CR 25 mg, but not with paroxetine CR 12.5 mg (remission: 41% with paroxetine CR 25 mg v
31% with paroxetine CR 12.5 mg v
28% with placebo; paroxetine CR 25 mg v
placebo, OR 1.83, 95% CI 1.17 to 2.87; P = 0.008; CR 12.5 mg v
placebo, reported as not significant; P value and absolute numbers not reported).
We found one systematic review (search date 2004, 10 RCTs, 548 people) that compared the effectiveness of drug treatments for people with psychotic depression.
The review found no significant difference for treatment of psychotic depression with amitriptyline compared with placebo (1 RCT; RR 8.40, 95% CI 0.50 to 147.87; P = 0.14).
We found one systematic review (search date 2004, 8 RCTs, 792 people with atypical depression), which compared the clinical effectiveness of drug treatments.
The review found that MAOIs significantly improved atypical depression compared with placebo (4 RCTs, 250 people; mean effect size 0.45, 95% CI 0.35 to 0.60). However, this review found an asymmetrical distribution for comparisons between MAOIs and placebo, possibly owing to one study that showed a very high response rate difference between phenelzine and placebo (25/30 [83%] with phenelzine v
5/26 [19%] with placebo). It must be noted that this high response rate difference does not correlate with a relatively large sample size. Furthermore, this finding is by contrast with the lower response rate differences in the other three studies comparing the efficacy of MAOIs and placebo in people with atypical depression (first study: 24/34 [71%] with phenelzine v
13/47 [28%] with placebo; second study: 12/17 [71%] with phenelzine v
7/24 [29%] with placebo; third study: 21/36 [58%] with phenelzine v
10/36 [28%] with placebo).