Antidepressant Discontinuation Research: Relapse, Withdrawal, and Better Trial Design

Evidence scope: Prepared for an October 30, 2026 editorial reference date using the dated primary sources linked below. This is not an exhaustive search through that date; later evidence, corrections, and regulatory decisions require a fresh check before publication. No claim of current approval or availability is implied.

Disclosure: This article is not diagnosis, treatment advice, or a substitute for individualized care. Do not change treatment based on this article.

Stopping an antidepressant is not one scientific question. Researchers may ask whether maintenance reduces depressive relapse, how withdrawal symptoms develop, which tapering strategies are tolerable, or how people make decisions after feeling well. A single study rarely answers all of these. Clear reporting should identify the question actually tested, especially when readers may be weighing a personal decision that involves their own episode history, side effects, support, and preferences.

The ANTLER primary trial report compared maintenance with tapering and discontinuation among adults in UK primary care who felt well enough to consider stopping. Participants had at least two previous depressive episodes or had used antidepressants for two years or longer. The study included people taking citalopram, fluoxetine, sertraline, or mirtazapine. It therefore informs a defined long-term-treatment population, not everyone who has ever received an antidepressant.

What the randomized comparison found

The reported proportion experiencing relapse by 52 weeks was 39 percent in the maintenance group and 56 percent in the discontinuation group. The higher relapse rate after assignment to discontinuation is an important finding. It supports discussing recurrence risk when considering a change. It does not establish that every individual should continue indefinitely, because group averages do not determine a person's preferences, tolerability, previous illness course, or likely outcome under a different stopping strategy.

Equally, the fact that some participants discontinued without a recorded relapse does not establish that stopping will be straightforward for a particular reader. A trial estimates outcomes under its own protocol. An individual plan requires assessment and monitoring that an article cannot supply. The site's questions before changing treatment can help organize a consultation, but this article does not recommend a taper speed, dose sequence, or decision to stop.

Why withdrawal belongs in the interpretation

Withdrawal symptoms can overlap with symptoms used to describe a return of illness, and both can occur after medication reduction. That overlap makes measurement and timing important. It should not be used to dismiss every post-discontinuation difficulty as relapse, nor to assume that every difficulty is withdrawal. Researchers need methods that characterize symptoms carefully and explain how outcomes were classified. The trial reported withdrawal-related outcomes as well as depressive and anxiety symptoms, which are relevant to interpretation.

A maintenance-versus-discontinuation trial is not automatically a comparison of all possible tapering methods. The duration and structure of the study's taper matter, and results should not be generalized to untested strategies without qualification. Future studies can compare taper approaches directly and examine whether additional support changes the balance of symptoms, completion, and recurrence. That is a different research task from simply demonstrating that continued treatment and stopping under one protocol lead to different average outcomes.

Blinding and expectations remain complicated

Matching placebo can strengthen a discontinuation study, but noticeable symptoms may lead participants to infer their assignment. Expectations, concern about recurrence, and actual changes in wellbeing can then interact. This does not invalidate randomization. It means readers should examine masking and symptom measurement rather than imagine that a double-blind label removes every interpretive problem. The most useful evidence acknowledges these challenges and explores their implications without overstating what can be resolved retrospectively.

The population also limits extrapolation. People who feel well enough to consider stopping differ from people currently experiencing a severe episode. The trial did not establish a stopping strategy for every medication, diagnosis, age group, or treatment duration. Our screening-versus-diagnosis article underscores why diagnostic context matters. Decisions involving bipolar disorder, for example, should not be inferred from a primary-care maintenance trial framed around this particular depression population.

Better trials need patient-valued outcomes

Relapse is important, but patients may also care about cognitive symptoms, emotional experience, sexual function, sleep, work, and the burden of ongoing treatment. A useful trial should make room for both benefits and burdens of maintenance and discontinuation. It should also track how many people resume treatment, why they do so, and how they fare afterward. A return to treatment may reflect several different experiences, which a single binary outcome cannot fully describe.

Future studies should report the support available during tapering and after emerging symptoms. They could evaluate structured follow-up, accessible clinical contact, and different approaches to managing uncertainty. These additions would help distinguish the effects of a medication change from the effects of being left without adequate support. Research should also explain how missing data and treatment changes are handled so that readers can understand whether results describe assignment, actual exposure, or a broader strategy of care.

A balanced interpretation for current conversations

ANTLER provides evidence that discontinuation under the studied conditions increased relapse risk over a year compared with maintenance. It does not settle every withdrawal question or dictate a lifelong decision. Those statements can both be true. Responsible reporting should resist polarized interpretations that turn the study into proof that everyone must stay on treatment or proof that every measured relapse was misclassified. Neither conclusion follows from the primary comparison alone.

A reader considering change should discuss it with the prescriber and agree how emerging symptoms will be assessed. The site's research-reading guide can help separate trial findings from broader claims. If there is sudden deterioration or an immediate safety concern, seek prompt local professional help. The future research priority is a more precise understanding of who can stop, under which conditions, with what support, and with which trade-offs—not a universal instruction derived from a group average.

Sources

• Maintenance or Discontinuation of Antidepressants in Primary Care

For general medicine-safety background, see NIMH mental health medications; this is background, not independent confirmation of this article’s specific study.