Almost all placebo-controlled drug trials in psychiatry are flawed due to their cold turkey design.1 Patients who are already in treatment first go through a “wash-out” period, when everyone is removed from the drug they had been on, and then randomized to placebo or drug. The wash-out period is too short to avoid abstinence symptoms in the placebo group.2
These abstinence symptoms can be the same as those that define the disorder, e.g. depression in depression trials, and the withdrawal effects can be very pronounced. One third of patients in long-term treatment with sertraline or paroxetine had an increase in their Hamilton score of at least 8 during a 5-8 day period where the drug was substituted with a placebo.3
We currently do systematic reviews of psychiatric drug trials at the Nordic Cochrane Centre where we only include trials in drug-naïve patients. The only such trial we found of neuroleptics in patients with psychosis was fraudulent (submitted manuscript).
In depression, trials exist in drug-naïve patients who have a chronic disease and developed depression in addition to this. An exemplary trial of this type was published in November 2017 in JAMA.4 It studied 201 patients with chronic kidney disease and depression for 12 weeks. Sertraline had no effect. The QIDS-C16 score changed by −4.1 in the sertraline group and by −4.2 in the placebo group; between-group difference, 0.1 (95% confidence interval −1.1 to 1.3; P = 0.82). The authors reported that a difference of 2 points on the QIDS-C16 is equivalent to a difference of 3 points on the Hamilton Rating Scale. The least clinically relevant effect on the Hamilton scale is 5-6.5
The authors hypothesized that the lack of effect of antidepressants — also in other trials in patients with chronic diseases — could be due to too low dose or to the presence of comorbidity.
I find both explanations unlikely. Receptor studies have shown that also when low doses are being used, the patients are dosed sufficiently because the dose-response curve has already flattened out.6DASB positron emission tomography study. Am J Psychiatry 2004;161:826-35.] In agreement with this, independently performed systematic reviews of dose-response studies of the effect in patients have been negative.7 The other argument is equally implausible. If depression pills work for depression, they should also work for depression in people with comorbidities.
A much more likely explanation is that the patients with chronic comorbidity in these trials were not already in treatment with depression pills before they were randomised.
Even helped by the withdrawal bias and bias caused by insufficient blinding, traditional trials of antidepressants have only found an effect of around 2 on the Hamilton scale.8
I have come to the conclusion that depression pills have no true effect on depression. What is being measured in most trials is likely bias. It is often claimed that they work for very severe depression but that isn’t correct either. The seemingly “better” effect (which is still clinically irrelevant) in very severe depression could simply be a mathematical artefact: the greater the score at baseline, the greater the bias will be when the “effect” is being assessed.9
I submitted a short letter to the editor of JAMA pointing out the likely explanation for the lack of effect of sertraline. I dutifully called depression pills antidepressants, although I prefer to avoid this term, as it is so misleading. Antibiotics can cure infections; depression pills cannot cure anything. JAMA rejected my letter, which “did not receive a high enough priority rating for publication in JAMA.” I have experienced this many times when I have submitted letters. Explaining why psychiatric drugs don’t work is not a priority for medical journals, it seems.