Stopping Antipsychotics: Relapse Risk and What the Trials Show

In first-episode trials, relapse was more than twice as frequent after stopping an antipsychotic as on maintenance. In 12 randomized trials with 1,133 adults stabilized after a first episode, 53.3% of those who discontinued relapsed within 12 months, compared with 21.0% of those who continued.¹ An earlier meta-analysis by the same group found that 45.7% of people who discontinued had not relapsed at 12 months, and 39.4% at 18 to 24 months.²

A Dutch trial's authors called for further research to find predictors of successful discontinuation.³ No tapering schedule appears here, because the decision and the schedule belong to the treating psychiatrist. NICE advises that if antipsychotic medication is withdrawn, it be done gradually, with regular monitoring for signs and symptoms of relapse.⁴

Relapse on maintenance compared with discontinuation

Across 65 placebo-controlled trials with 6,493 patients with schizophrenia, 27% relapsed within a year on an antipsychotic and 64% on placebo.⁶ A 2020 Cochrane review of 75 trials found relapse in 24% and 61% at 7 to 12 months.⁷

In the Cochrane review, at least one movement disorder occurred in 14% on antipsychotics and 8% on placebo, and weight gain in 9% and 6%.⁷

The difference between drug and placebo decreased with study length, and the Cochrane authors limited their conclusion to approximately two years of follow-up.^{6, 7}

The dose-reduction trial and its seven-year follow-up

A Dutch trial randomized 131 people whose first episode had been in remission for 6 months to guided dose reduction and discontinuation or to maintenance. Over 18 months, relapse occurred in 43% with the discontinuation strategy and 21% with maintenance. About 20% were successfully discontinued, about 30% restarted because symptoms recurred, and functional outcome showed no advantage.³

Seven years later, 103 of 128 patients were assessed. Recovery was recorded in 40.4% of the dose-reduction group and 17.6% of the maintenance group. Treatment after the trial was at the clinician's discretion, and the authors wrote that more studies were necessary before the results were incorporated into general practice.⁸

The 10-year Hong Kong follow-up

A Hong Kong trial assigned 178 people, free of positive symptoms after at least a year of treatment, to quetiapine or placebo for 12 months. The estimated risk of relapse at 12 months was 41% on quetiapine and 79% on placebo.⁹ Discontinuation here meant a switch to placebo, and the abstract describes no taper.

At 10 years, a poor outcome (persistent psychotic symptoms, a need for clozapine, or death by suicide) had occurred in 39% of the discontinuation group and 21% of the maintenance group, a risk ratio of 1.84. Four people in the discontinuation group and two in the maintenance group died by suicide during follow-up.¹⁰

Two recent trials of gradual reduction

Sources: journal abstracts for both trials.^{11, 12} The RADAR severe-relapse figures come from the funder's 2026 programme report and do not appear in the journal abstract.¹³

The HAMLETT authors noted that doses were comparable in the two groups from one year onward, so the later difference in functioning "is not a direct result of lower medication."¹¹ In RADAR the median dose reduction was 67% at any point and 33% at 24 months.¹²

The TAILOR trial was terminated for insufficient recruitment after 29 participants and reached no conclusion.¹⁴ A Cochrane review of 25 dose-reduction trials found no difference in quality of life or functioning and a higher risk of relapse (risk ratio 2.16, low-certainty evidence).¹⁵

"Gradual" has no single definition

A meta-analysis of 5 studies with 261 people found withdrawal symptoms in 53% after discontinuation, and its authors concluded that they appear to occur frequently after abrupt discontinuation.⁵

In the 2019 first-episode meta-analysis, four trials stopped the drug abruptly and six tapered over a mean of 3 months.² In the Cochrane dose-reduction review, gradual meant 2 to 16 weeks.¹⁵ In HAMLETT, tapering lasted 124 days on average, with a range of 6 to 334 days.¹⁶ No schedule is given here, because the decision and the schedule for a given patient belong to the treating psychiatrist.

Whether a slower taper lowers relapse risk is unsettled. Among 227 HAMLETT participants who tapered, 45.8% relapsed in the first 18 months after remission, and tapering speed did not predict relapse.¹⁶

Antipsychotic use and mortality

In Finnish registers covering 66,881 patients with schizophrenia, 7 to 11 years of cumulative antipsychotic exposure was associated with lower mortality than no drug use (hazard ratio 0.81).¹⁷ In a later Finnish cohort of 62,250, the adjusted hazard ratio for death from any cause during antipsychotic use compared with non-use was 0.48.¹⁸

These findings are observational, compare different people, and cannot establish a causal relationship. The randomized trials reported too few deaths to test the question.⁶

What the guidelines say about how long to continue

NICE says to inform patients that "there is a high risk of relapse if they stop medication in the next 1 to 2 years" and to keep monitoring for at least 2 years after withdrawal.⁴ The 2023 VA/DoD guideline recommends maintenance treatment for people who have responded and states that the evidence "precludes a treatment recommendation for the optimal length of treatment of a first episode of psychosis."²⁰ In the text reviewed here, neither sets a minimum number of months or years.

At the 20-year follow-up of the OPUS cohort, 36% of the 143 people interviewed were in remission off medication. Attrition was 71%, which the authors said may inflate favorable results, and they noted that those still in treatment likely had more severe illness.¹⁹ I've written separately about first-episode psychosis treatment.

Limitations of the evidence

This evidence has limitations. First, most relapse comparisons run 1 to 2 years. Second, the long-term follow-ups disagree, with better recovery after early dose reduction in the Dutch trial and worse outcomes after early discontinuation in Hong Kong. Nevertheless, relapse was consistently more frequent after stopping or reducing. A change in dose is a decision to make with the treating psychiatrist, with monitoring for relapse planned in advance.

In a crisis: call or text 988, or go to the nearest emergency department. A first episode of psychosis belongs in a hospital-based or coordinated specialty care program.

I start and manage clozapine in an outpatient practice in Chelsea, Manhattan. Details are on the schizophrenia care in NYC page. Clozapine should not be stopped, skipped, or restarted without the prescriber. This page is general information and does not replace an evaluation.

Related: Schizophrenia and psychosis treatment guide · When to start clozapine · Clozapine blood tests after the REMS ended · Clozapine side effects · Missed clozapine doses · Clozapine and suicide risk · Cannabis and psychosis · Schizoaffective vs schizophrenia · Cobenfy vs clozapine · Postpartum psychosis · Injections vs pills · Stimulants and psychosis · Drug-induced psychosis · Clozapine levels · CBT for psychosis · Clozapine myocarditis · Cannabis after psychosis · Clozapine constipation · Work and school · Clozapine drooling · Early signs in teens · Clozapine weight gain · Genetic and family risk · Who can prescribe clozapine · Schizophrenia and violence · Clozapine in teenagers · Sleepiness, bedwetting, heart rate · When clozapine isn't enough · Clozapine interactions · Benign ethnic neutropenia · Clozapine off-label uses · Starting clozapine · Clozapine and pneumonia · Clozapine seizures · Clozapine in pregnancy · Clozapine rechallenge · Relative refuses treatment · Late-onset schizophrenia · Negative symptoms · First-episode psychosis · Tardive dyskinesia · Schizophrenia life expectancy · Key clozapine studies