Monday, September 23, 2019

Scientists think antidepressants work but is the evidence biased?

Gemma Lewis and Glyn Lewis popularise their PANDA study (published in The Lancet Psychiatry) in The Conversation. The headline of their piece is 'Antidepressants work, but just not how scientists thought they worked'.

Even though the PANDA study did not find a significant difference in depression scores at 6 weeks between sertraline and placebo groups, the authors still claim antidepressants work. This is because they found evidence of a significant difference for anxiety symptoms and suggest the benefit of antidepressants is more for anxiety than depression. I'm not sure if they're saying that antidepressants should be renamed anxiolytics.

As I've pointed out previously (eg. see post), despite antidepressant trials on average showing a statistically significant difference for active drug over placebo, the difference is small and there is a question about how clinically significant this difference is. Furthermore, methodological difficulties, such as unblinding, can bias the results, so it is possible that any statistically significant result is an artefact (see my Bias in controlled trials webpage). This is called the placebo amplification hypothesis of the apparent statistical advantage of antidepressants over placebo in clinical trials. It is difficult to prove and the debate about antidepressant efficacy is still open in the literature (see previous post), despite the Lewises apparently not being prepared to admit that their PANDA study may actually provide evidence that antidepressants are not effective.

Data is given on unblinding in the PANDA study, although the authors do not make very much of it. The authors knew of three incidents when participants opened the capsule to see if there was a tablet included, and these patients were withdrawn from the trial. The majority of participants did not think they were on active treatment, even though half of them were. More people on sertraline (46%) thought they were on active treatment than those on placebo (19%). People seem to have generally thought they were not on active treatment, and placebo patients were quite good at recognising this. Participants were able, therefore, for whatever reason, to distinguish sertraline from placebo, so it's misleading to say the PANDA trial was double-blind.

Despite what the authors seem to think, antidepressant trials are not adequately blinded (Even et al, 2000). The findings of the PANDA study may therefore merely reflect the authors bias (transmitted to the participants) that antidepressants are effective (although for some reason not detected at 6 weeks with depression scores). Antidepressants may merely be placebo panaceas for emotional problems.