Thread by @markhoro

Startling lack of interest in confounding by withdrawal symptoms in this antidepressant discontinuation study in NEJM: bit.ly/3AYzP2Z Withdrawal symptoms in this discontinued group render this study uninterpretable. (1/n)

Withdrawal symptoms in the discontinuation group would have inflated the apparent rate of relapse in this group (2/n)

Patients on ADs for at least 9 months were taken off by halving for a month and then halving further for a month before stopping. Much quicker than the ‘months or longer’ recommended by RCPsych guidance and much faster than most patients can tolerate.bit.ly/39QwOpB (3/n)

This caused withdrawal symptoms esp. at week 12. They measured modified withdrawal scale (15 symptoms) and found on average 3.1 in discon group vs 1.3 in maintenance group (but no measure of severity) (4/n)

Withdrawal symptoms include depressed mood, anxiety, anhedonia, disrupted sleep, fatigue and concentration problems. The rCIS-R measured depressed mood, anhedonia, fatigue, loss of concentration and sleep disturbance. (5/n)

Therefore withdrawal symptoms would have registered on the rCIS-R and increased the chance of registering as relapse. (6/n)

Drug doses were halved weeks 0-4 which has almost no effect on their receptor occupancy – no excess relapses were detected. Halved again (by second daily dosing) from weeks 4 to8 with some increase in relapses in the discon group. (7/n)

Drugs stopped at week 8. Almost entire separation between the lines occurred in the following 6-12 weeks the time at which withdrawal effects are most likely. (8/n)

At end of trial anxiety scores (GAD-7) were the same between maintenance and discon groups (3.0 vs 3.1, respectively) but they had been very different at week 12 (3.1 vs 5.3) the time at which withdrawal effects would have been most pronounced (4 weeks after stopping) (9/n)

The same was true for depression scores (PHQ-9). At 12 weeks maintenance versus discon was 4.1 vs 6.3 but by weeks 52 it was very similar 3.7 vs 4.0 (10/n)

Partial explanation for this was that patients who discontinued returned to their drugs but at end of trial only 44% of discon group was on ADs, while 89% in maintenance group were and they still had the same anxiety and depression levels. (11/n)

The pattern of anxiety and depression scores matches that of withdrawal effects: spiked at week 12, improving in weeks 24 through to 52. Differences between groups in PHQ-9 for week 12, 24, 39 and 52: 2,2, 0.8, 0.6, 0.3. For withdrawal symptoms: 1.8, 0.5, 0.9, 0.3. (12/n)

For GAD-7: 2.2, 0.7, 0.9, 0.1. (13/n)

Occam’s razor suggests rather than having 2 things going on: relapse and withdrawal symptoms it is more likely that increased anxiety and depression that correlate with the (mostly physical) withdrawal symptoms measured by the withdrawal scale is part of withdrawal syndrome(13/n)

If patients who scored significantly on withdrawal scale were excluded from registering as relapse then it is likely that the difference between rate of relapse in the discontinuation arm and the maintenance arm would be much less (14/n)

The failure to take into account mis-diagnosing withdrawal symptoms as relapse makes it difficult to interpret this trial and conclude as the authors do that people who stop antidepressants relapse more than those who do not. (15/n)

Would be similar to concluding that people who stop benzos have a relapse of anxiety more than those who continue (when anxiety is clearly a withdrawal symptom) or that people who quit smoking should continue in order to treat the anxiety that increases when they stop (16/n)

Ignoring confounding by withdrawal effects makes most discontinuation studies in psychiatry ‘uninterpretable’ to quote a very polite paper on the topic: bit.ly/3oi7cdx And further analysis: bit.ly/3oivX9F (16/n)

To properly test the relapse prevention properties of antidepressants either very gradual tapering that prevents withdrawal symtoms from arising is needed or patients who experience significant withdrawal symptoms should be excluded from registering as relapse (17/n)

The media articles leaping to the conclusion that this study shows that people should be on these drugs long-term are very misleading to the public and are going to contribute to the ongoing and unnecessary use of these medications. (18/18)