Stopping an antipsychotic: why a quick relapse is usually the illness, not "withdrawal"

Research digest · Written by Dr. Saurabh Kumar · 14 October 2026

The question I hear in many family conversations

A father leans forward and says, "Doctor, he has been well for two years now. Can we stop the medicine?" Behind that question sits a quieter fear. If we stop and he becomes unwell within days, was it the illness coming back, or was it the body reacting to the missing tablet?

That second question is not academic. If a fast relapse is a withdrawal reaction, the answer is to taper more slowly next time. If it is the illness itself, the answer is different: know who is at risk, watch closely, and act early. A new analysis from the University of Oxford and King's College London, published in The Lancet Psychiatry this month, tried to settle it using the best data available.¹

Two explanations that have argued for decades

1. 1The illness returns

   Antipsychotics work mainly by dampening dopamine signalling. Once that effect falls below a threshold, symptoms that were always under the surface come back. On this view, the speed of stopping matters little.

2. 2Withdrawal

   Long periods of dampening may make the brain's dopamine receptors oversensitive. Remove the medicine and the system can rebound, sometimes called "supersensitivity psychosis". On this view, very slow tapering should protect people.

Both ideas have supporters and some laboratory backing. What was missing was a direct test in real trial data.

What the researchers did

The team went to the Yale Open Data Access project, which holds participant-level records of trials that drug companies have agreed to share.³ They found five double-blind trials of one second-generation antipsychotic, given as tablets or as a long-acting injection. Adults with schizophrenia or schizoaffective disorder had first been stabilised on the medicine for more than three months. They were then randomly assigned, without anyone knowing which group they were in, to continue the medicine or switch to a placebo. That switch is, in effect, a sudden stop.

The researchers looked only at the 417 people who went on to relapse, 271 from the injection trials and 146 from the tablet trials, and traced their symptom scores visit by visit up to relapse. A statistical method called latent class mixed modelling let the data reveal its own patterns rather than imposing one.²

What they found

Two patterns emerged in both sets of trials. A smaller group relapsed rapidly, with symptoms already fluctuating from the start. A larger group relapsed later, after a longer period of relative stability.

Then came the decisive test. If withdrawal drives rapid relapse, the rapid pattern should be far more common among people who had stopped the medicine. It was not. In the tablet trials the share of rapid relapses was almost identical whether people had stopped (about 27%) or continued (about 26%). In the injection trials it was higher in those who stopped (about 20% versus 11%), but the difference was not statistically significant and could be chance.

Two more findings point the same way. People who stopped showed no distinct withdrawal signature at relapse: anxiety, agitation and low mood were no more prominent than in people who relapsed while still taking the medicine. And people who relapsed fast were already more symptomatic at the start of the trial, before any randomisation. Rapid relapse was associated with pre-existing vulnerability rather than with the act of stopping.

What this does and does not change

This study does not say that stopping is safe. Earlier meta-analyses have found that relapse is roughly twice as likely after stopping an antipsychotic than with continued treatment, and in these trials about three-quarters of all relapses occurred in people who had stopped. That part of the picture is unchanged.

What the study shifts is how we read a fast relapse, and what we rely on to prevent one. Its findings fit with other recent trials suggesting that tapering speed, by itself, may not change relapse risk. A gradual reduction is still sensible, because it gives everyone time to notice early warning signs. But it should not be mistaken for a safety net.

For families, the practical points are these:

• A sudden relapse after stopping is not proof that the taper was done wrong, and it is not the family's fault. It is more likely a sign that the illness remains active.
• People most likely to relapse quickly were often not fully stable to begin with. Baseline stability, past relapses and past response to treatment should weigh more in the decision than any fixed tapering schedule.
• Any plan to reduce or stop should include a written list of that person's own early warning signs, a named family member who watches for them, and a clear route back to the treating doctor within days, not weeks.
• Stopping is often decided for reasons unrelated to the illness, such as cost, side effects or a belief that the medicine is habit-forming. This evidence shifts the concern: the main risk is the illness returning, and sometimes it returns fast.

Evidence grade

Limits

The analysis covers one antipsychotic and its paliperidone formulations. Whether the same two patterns appear after stopping other antipsychotics, including those used in treatment-resistant illness, is not known. The authors note that withdrawal-type relapse may still occur in specific situations, such as abrupt stopping after long high-dose treatment. The trials were not designed to measure withdrawal, so the absence of a withdrawal signature is an indirect inference from symptom scales. Trial participants tend to be more stable and more adherent than people in routine clinical care, and none of the trials were run in India. The method identifies patterns in data rather than causes, and the finding has not yet been replicated by an independent group.

Frequently Asked Questions

• Can antipsychotics be stopped after two years of being well?

  Sometimes, for some people. The decision depends on how many episodes there have been, how stable the person is now, and how any past relapse unfolded. It should be made with the treating psychiatrist, not alone.

• What happens if antipsychotic medication is stopped suddenly?

  Relapse risk rises; across trials it is roughly double that of people who continue. Some people relapse within days, most within months, and some not at all. This analysis suggests that a fast relapse is usually the illness returning, not a withdrawal effect.

• Is "supersensitivity psychosis" real?

  It is a real hypothesis with laboratory support. This analysis did not find evidence that it explains a meaningful share of relapses in trial data, though the authors do not rule it out in particular circumstances.

• Does slow tapering prevent relapse in schizophrenia?

  Current evidence, including this study, suggests tapering speed on its own may not change relapse risk. Slow reduction is still reasonable because it gives time to notice early signs and respond.

• What are early warning signs of a psychotic relapse?

  These are personal, but families often notice changes in sleep, withdrawal from people, irritability, suspiciousness, or a return of the thoughts or experiences that marked earlier episodes. Families often recognise the pattern before the person does, which is why a written list helps.

• Who is most likely to relapse quickly after stopping?

  In this analysis, people whose symptoms were already higher at the start of the trial. In practice, stability should be judged honestly before any reduction is planned.

• My relative stopped the medicine on his own and is unwell again. What should we do?

  Contact the treating psychiatrist as soon as possible. If there is any risk to safety, go to the nearest emergency service or call 112.

• Is this study relevant to patients in India?

  The biology of relapse is unlikely to differ, but the trials were not run in India and participants were more stable than many people seen in routine care here. Treat the findings as useful guidance, not as a rule.

About this article

Written and medically reviewed by

Dr. Saurabh Kumar

• MD (Psychiatry) AIIMS New Delhi
• Associate Professor & HOD, Dept. of Psychiatry, RDJM Medical College, Muzaffarpur
• Ex-Assistant Professor, PMCH Patna
• Consultant, Healthy Minds India, Patna

Written

14 October 2026

Format

Research digest

Evidence

Individual participant data meta-analysis of randomised trials

Sources

3 references

• Transparency: this article was prepared with AI assistance and reviewed and signed off by Dr. Saurabh Kumar.

© 2026 Healthy Minds India, Patna · Research digest, October 2026