Stopping antidepressants: how common are discontinuation symptoms, and what does tapering change?

In 2025, three meta-analyses of symptoms after stopping antidepressants used different methods and reached different estimates ^[1][2][3]. In 2026, a network meta-analysis of 76 trials compared ways to stop by later relapse in remitted depression or anxiety ^[4].

What changed in 2025 and 2026?

• JAMA Psychiatry, September 2025: in the Kalfas meta-analysis of 50 randomized trials (17,828 participants), stopping raised symptom scores at week 1 (standardized mean difference [SMD] 0.31, 95% CI 0.23 to 0.39), equivalent to about one more symptom on the Discontinuation-Emergent Signs and Symptoms scale ^[1].
• Molecular Psychiatry, May 2025: the Zhang meta-analysis of 35 studies pooled the incidence of withdrawal symptoms at 42.9% ^[2].
• Psychological Medicine, July 2025: the Moncrieff reanalysis of the 2024 Henssler review ^[5] used five studies with systematic measures and pooled an incidence of 55% (95% CI 31% to 81%; n=601) without subtracting nocebo effects ^[3].
• Lancet Psychiatry, January 2026: the Zaccoletti network meta-analysis of 76 trials (17,379 participants) compared abrupt stops, fast and slow tapers, lower doses and continuation, each with or without psychological support ^[4].

How often do discontinuation symptoms occur, and how do they differ from relapse?

In the 2024 Lancet Psychiatry meta-analysis of 79 studies (21,002 patients), at least one symptom occurred in 31% after an antidepressant was stopped and in 17% after placebo ^[5]. The 2024 authors allowed for these nonspecific effects and put the incidence at about 15%, with severe symptoms at 2.8% versus 0.6% after placebo ^[5]. The 2025 Psychological Medicine reanalysis argued that the 2024 estimate rested mostly on adverse events reported in passing in industry efficacy trials ^[3]. In the 2025 Molecular Psychiatry meta-analysis, incidence rose with treatment duration, from 35.1% after 6 to 12 weeks to 51.4% after more than 24 weeks ^[2]. The 2025 meta-analysis also reported 34.5% incidence after tapering and 42.5% after abrupt cessation, a difference that was not statistically significant ^[2].

In the 2025 JAMA Psychiatry meta-analysis, stopping raised the odds of dizziness (OR 5.52, 95% CI 3.81 to 8.01), nausea (OR 3.16, 95% CI 2.01 to 4.96), vertigo (OR 6.40, 95% CI 1.20 to 34.19) and nervousness (OR 3.15, 95% CI 1.29 to 7.64) compared with stopping placebo ^[1]. The same analysis found no rise in depression symptoms in five studies of major depression, and its authors interpreted later low mood as relapse rather than withdrawal ^[1]. The 2025 Psychological Medicine reanalysis listed potential misclassification of withdrawal symptoms as relapse among the shortcomings of the 2024 meta-analysis ^[3].

How do tapering and continuation compare for relapse risk?

In the 2026 network meta-analysis, relapse risk against an abrupt stop was lower with continuation plus psychological support (RR 0.40, 95% CI 0.26 to 0.61), continuation alone (RR 0.51, 95% CI 0.46 to 0.58), a taper over more than 4 weeks plus psychological support (RR 0.52, 95% CI 0.38 to 0.72) and continuation at a reduced dose (RR 0.62, 95% CI 0.42 to 0.92) ^[4]. A slow taper alone showed no clear gain over an abrupt stop (RR 0.81, 95% CI 0.56 to 1.18; low certainty) ^[4]. In a 2021 primary-care trial of 478 adults, relapse by 52 weeks occurred in 56% of those randomized to taper and discontinue and in 39% of those who maintained treatment (HR 2.06, 95% CI 1.56 to 2.70) ^[6]. In obsessive-compulsive disorder (OCD), a 2025 meta-analysis of 9 trials (n=1,084) found a relapse risk ratio of 0.53 (95% CI 0.42 to 0.68) with maintenance ^[7].

Where is the evidence uncertain?

In the 2026 network meta-analysis, most comparisons drew on depression trials, so evidence in remitted anxiety disorders is limited ^[4]. In the 2025 Psychological Medicine reanalysis, prior treatment lasted 12 weeks or less in four of the five included studies, and heterogeneity was high ^[3]. The 2025 Molecular Psychiatry authors call for discontinuation studies in long-term users with extended follow-up ^[2].

At the point of care: what to check

• Duration of prior antidepressant use, since withdrawal incidence rose with treatment length in the 2025 Molecular Psychiatry meta-analysis ^[2].
• The type and timing of new symptoms; dizziness, nausea, vertigo and nervousness had raised odds after stopping versus placebo in the 2025 JAMA Psychiatry meta-analysis ^[1].
• Timing of mood worsening, since the 2025 JAMA Psychiatry authors read later low mood as relapse, while the 2025 reanalysis flagged possible misclassification of withdrawal as relapse in the 2024 meta-analysis ^[1][3].
• Taper speed and whether psychological support is available, since slow tapering with support lowered relapse risk against abrupt discontinuation while slow tapering alone did not ^[4].
• The treated disorder, since OCD has its own maintenance-versus-discontinuation relapse estimate ^[7].