In a systematic review and meta-analysis, Kalfas et el (2025) conclude that the mean number of discontinuation symptoms at week 1 after stopping antidepressants was below the threshold for clinically significant discontinuation syndrome. The article is motivated to show that antidepressant withdrawal may be less common that is thought to be the case (see New Scientist article). The difficulty in obtaining estimates of incidence is made worse by the lack of methodologically rigorous, randomised placebo-controlled trials in real world settings, certainly over the longer-term rather than just after one week. In fact, initially on stopping antidepressants, people can be hopeful that they no longer need antidepressants.
The data including in the meta-analysis was from 11 trials with short duration of use of antidepressant: 6 for 8 weeks; 4 for 12 weeks; and one for 26 weeks. People can stay on antidepressants for considerable periods of time and the risk of withdrawal symptoms increases with duration of use (see eg. Horowitz et al, 2025). The data obtained after one week discontinuation may well not be generalisable to the bulk of long-term users, particularly if followed up for more than a week.
The Telegraph reports that the research team have suggested guidelines need to be re-written to reassure people that they are unlikely to experience severe side effects when coming off antidepressants, which is clearly unjustified (see eg. article by Mark Horowitz & Joanna Moncrieff on The Conversation). Such advice is contrary to current guidance that people should be informed of the risk of withdrawal symptoms on first starting antidepressants (see eg. previous post). Doctors should be encouraged to follow this guidance, rather than the proposal by Kalfas et al. At least the researchers have subsequently admitted that most studies they reviewed had only followed up patients for two weeks and that antidepressant treatments they studied were shorter than those commonly prescribed in the real world, although this isn’t as explicit in the paper as it should be.