Read on Twitter
I am still a bit puzzled, how fiercly some psychiatrists debated the conclusion of my recent paper, that in depression relapse prevention trials preventive effects and withdrawal effects are severely confounded. So I checked various relapse prevention trials for anxiety disorders
Consistent with excess risk in the first few weeks and no preventive effects after 12 weeks in depression trials, I found exactly the same pattern in discontinuation trials for anxiety disorders. That is, drug-placebo separation only at the beginning; no preventive effects later
Here is an example from a escitalopram relapse prevention trial for social anxiety disorder: https://www.ncbi.nlm.nih.gov/pubmed/16259540
High risk of relapse during the first 6-8 weeks after discontinuation but almost no drug-placebo sepration, that is, preventive effects, afterwards
High risk of relapse during the first 6-8 weeks after discontinuation but almost no drug-placebo sepration, that is, preventive effects, afterwards
What we clearly see here is that, relative to placebo, escitalopram prevents against "relapse" only during the first 30-40 days, afterwards the curves run almost perfectly parallel, which means that there was no more relapse prevention on escitalopram than on placebo later.
Why does escitalopram prevent relapses only for the first 30-40 days? Well, IMO it doesn't prevent relapse but rather withdrawal reactions. So was there an excess of withdrawal reactions immediately after discontinuation in the placebo arm (as assessed with DESS)? Yes there was:
"One week after randomization, 9% of the patients in the escitalopram group had a total DESS score of ≥4, compared to 27% of the patients in the placebo group (p < .001)." That is, symptoms commonly part of a withdrawal syndrome were found in 9% on active vs 27% on placebo
If you deny that this is withdrawal confounding at its best, then I'm really baffled. Can anyone really state with confidence that there is no withdrawal confounding in such trials? How can't the authors see that the drug-placebo separation is best explained by withdrawal?
You think I cherry-pick? Here is another relapse prevention trial. This one tested duloxetine for relapse prevention in generalized anxiety disorder: https://www.ncbi.nlm.nih.gov/pubmed/18559291
Do the data again confirm that drug-placebo separation occurs almost entirely in the first few weeks?
Do the data again confirm that drug-placebo separation occurs almost entirely in the first few weeks?
Yes, again they do. Here we see anxiety scores before and after randomization. See how anxiety increases shortly of discontinuation in the placebo group, but no preventive effect for duloxetine relative to placebo afterwards:
Withdrawal symptoms were not systematically assessed in this trial. Only spontaneous reports, but still: "Compared with patients in the duloxetine group, dizziness was the only adverse event to occur significantly more often in the placebo group (9.9% vs 3.7%, P ≤ 0.05)."
Other data are not reported, but we all know that if we rely on statistically significant differences only, we miss many important adverse events that may considerably differ between groups. Yet, higher incidence of dizziness on duloxetine clearly shows that withdrawal occurred
The data again reveal that what is assumed to be a preventive effect is present only shortly after discontinuation. After a few weeks, duloxetine has no effect relative to placebo. IMO this again strongly supports the view that the drug prevents withdrawal rather than relapse
In conclusion, I have now scrutinized dozens of relapse prevention trials, and I found the same pattern strongly suggestive of withdrawal confounding in every signle one of them. The data are just not compatible with the common view that drug continuation prevents true relapse
