The American FDA has reported on 228 trials of antidepressants versus placebo that were submitted compulsorily to them (thus avoiding publication bias).1 Response to both placebo and active medication increased with the severity of depression, with the difference between medication and placebo being largest with the most severe depression.
In a JAMA review, the number needed to treat reported for patients with severe depression was 4, 11 for moderate and 16 for mild depression.2 The NNT of 16 means that for every 100 patients given antidepressants, 40 will improve with placebo alone, 6 improve from the medication, and 54 will not improve. Even in severe depression, more patients respond to the placebo, about 40%, than to medication at 25%.3
Many patients with depression, even severe depression, will improve in a few weeks; so talk first and prescribe later, unless strong contextual reasons prevail.4