Going slower is better.
Patients’ risks for relapse increase when lithium or antipsychotic medications are discontinued rapidly rather than gradually. To compare rapid (1–7 days) versus slow (14 days) discontinuation of antidepressants, researchers in Sardinia followed 398 patients with recurrent major depressive disorder (n=224), panic disorder (n=75), bipolar II disorder (n=62), or bipolar I disorder (n=37). Follow-up lasted at least 1 year (mean, 2.8 years; mean length of antidepressant treatment, 8.5 months).
In this observational study, the treating clinicians or the patients had chosen to discontinue medications when patients were clinically well; antidepressants were withdrawn rapidly in 188 patients and gradually in 210. Rapid discontinuation was associated with a significantly shorter time to relapse than gradual discontinuation (median, 3.6 vs. 8.4 months). The authors estimated that time to relapse after rapid discontinuation was only 25% of the average time to earlier relapses in the same patients. The findings were most evident for bipolar I and panic disorders. After rapid discontinuation, the intervals preceding relapse were similar among different drug types, but after gradual discontinuation, time to relapse was substantially longer with tricyclics than with modern antidepressants. Relapse risk was the least pronounced for medications with prolonged half-lives and was not associated with antidepressant dose, duration of illness or treatment, or concurrent treatment.
Comment: Several uncertainties remain, and the study needs replication. Still, these findings align with study results on other psychotropics. Rapid discontinuation substantially increases relapse risks and shortens relapse latencies. Clinicians opting to discontinue antidepressants, particularly short-acting serotonin reuptake inhibitors, should plan to taper over weeks to months. As an editorialist notes, clinicians should forewarn patients who decide to abruptly discontinue their medications, especially pregnant women, of these risks.
