Rapid Relapse After Stopping Antipsychotics: Withdrawal or Relapse?
By Dr. Anindo Mitra, MBBS, MD Psychiatry (JIPMER), Consultant Psychiatrist, Gurugram
TL;DR
A 2026 Lancet Psychiatry analysis of five paliperidone trials studied 417 people whose illness relapsed, to ask whether fast relapse is more common after stopping medication than during continued treatment.
Among relapsers, the rapid pattern was not significantly more common after discontinuation: long-acting injectable (LAI) 20% (39/197) versus 11% (8/74), oral 27% (29/108) versus 26% (10/38).
Higher baseline symptom severity was the clearest predictor of the rapid pattern.
This does not show that stopping is safe. Stopping an antipsychotic still raises overall relapse risk, and the study covers one drug.
What is rapid relapse after stopping an antipsychotic?
Rapid relapse means psychotic symptoms worsen soon after an antipsychotic is reduced or stopped. In the 2026 analysis, the rapid group had a median time to relapse of 56 days on long-acting injectable paliperidone and 10 days on oral paliperidone. It describes a pattern of symptom change, not a separate diagnosis.
The word “rapid” is a label the researchers derived from the data within each trial group. The timing differs by formulation, so it should not be read as one universal cutoff.
Why does stopping an antipsychotic raise relapse risk at all?
Antipsychotics protect against relapse, so removing them removes that protection. A Lancet meta-analysis of 65 trials found that 27% of people on drug relapsed at 7 to 12 months, compared with 64% on placebo. That gap exists whatever the speed of relapse.
This matters because the 2026 paper does not challenge that finding. Its own framing accepts that discontinuation increases overall relapse risk. An earlier individual participant analysis of paliperidone trials also found that higher Clinical Global Impression severity at baseline was linked to greater risk after discontinuation, so severity was already a known signal.
What is antipsychotic withdrawal, and how is it different from relapse?
Withdrawal refers to new symptoms produced by the body adjusting after a drug is removed, while relapse is the return of the underlying illness. The two can overlap in timing and in features such as agitation, anxiety and insomnia, which is why separating them at the bedside is difficult.
Withdrawal is not a fringe idea. A systematic review concluded that withdrawal symptoms appear to occur frequently after abrupt discontinuation of oral antipsychotics. A separate meta-analysis found that new psychiatric adverse events were more frequent after discontinuation (odds ratio 2.01), and that tapered discontinuation can lower the risk of new somatic adverse events.
A further proposal, discussed by clinicians, is that abrupt removal of dopamine blockade in a vulnerable person could itself trigger a psychotic episode. That idea is why researchers wanted a direct test.
What did the 2026 Lancet Psychiatry study actually test?
Louie and colleagues pooled individual participant data from five paliperidone relapse-prevention trials, three long-acting injectable and two oral, and modelled symptom scores for the 417 participants who relapsed. They asked whether rapid worsening was over-represented after discontinuation compared with relapse during continued treatment.
That comparison is the logical test of a withdrawal explanation. If stopping caused a distinct rapid syndrome, rapid trajectories should cluster in the discontinuation arm. The team used latent class mixed modelling on repeated PANSS scores from randomisation to relapse, and an earlier preprint version described it as a test of the antipsychotic withdrawal syndrome hypothesis of relapse.
The analysis was not preregistered, and the authors report no lived-experience involvement in the analysis or writing.
What did the researchers find?
Two trajectories appeared in each formulation: rapid and delayed. Among relapsers, the rapid class made up 20% (39/197) of discontinuation-arm relapses versus 11% (8/74) of continued-treatment relapses for LAI (p=0.12), and 27% (29/108) versus 26% (10/38) for oral (p=0.95). Neither difference reached significance.
Among people who relapsed, time to relapse also did not differ by arm. For LAI it was 85.0 days after discontinuation and 81.5 days on active treatment. For oral it was 25.5 days in both groups. Symptom profiles at relapse did not differ by treatment assignment either, including agitation, anxiety and depression, which some have proposed as withdrawal indicators.
Delayed trajectories made up most relapses: 83% (224/271) for LAI and 73% (107/146) for oral.
Why does the denominator matter?
These percentages describe only people who relapsed: 417 across five trials. They are not the share of everyone who stopped medication, and they do not show how many people in each arm relapsed overall. Mixing those two denominators is how a careful finding turns into an inaccurate headline.
Consider what the numbers do not say. The 20% and 11% are proportions within two groups of relapsers, 197 and 74 people. They say that among those who relapsed, the fast pattern was not clearly more common after stopping. They do not say that stopping was no more likely to cause relapse, because overall relapse risk is higher after discontinuation and this analysis was designed around relapse itself.
What predicted a rapid trajectory?
Baseline symptom severity was the clearest predictor. Mean baseline PANSS total was 64.9 in the rapid LAI class versus 52.7 in the delayed class, and 58.6 versus 50.8 for oral, both p<0.001. Higher baseline Clinical Global Impression severity pointed the same way in both cohorts.
The authors interpret this as pre-existing illness susceptibility rather than a distinct pharmacological discontinuation effect. They also raise one possible measurement explanation: ratings may be held artificially low at screening to meet stability criteria, then drift back toward a person's usual level after randomisation. That is a hypothesis consistent with some patterns, not something the analysis demonstrated.
What can this study not tell us?
It cannot tell us that stopping an antipsychotic is safe, that relapse risk is unchanged, or that taper speed is irrelevant. A nonsignificant result is not proof of no difference, and the LAI odds ratio of 2.04 has a 95% confidence interval of 0.90 to 4.59, which leaves real uncertainty.
Several boundaries apply. All five trials studied paliperidone, so the findings may not extend to other antipsychotics or to abrupt clozapine cessation. Trial participants tend to be more stable and adherent than people in routine care. The trials were not designed to measure withdrawal, and PANSS is not a validated withdrawal scale, so experiences it does not capture may be missed. The authors also lacked data on prior relapse history and earlier antipsychotic exposure.
How should a rapid decline after stopping be interpreted?
A rapid decline after stopping should be assessed as possible recurrence of the underlying illness, while still considering withdrawal symptoms and other causes. In these trial populations, rapid worsening was not clearly specific to discontinuation, so the whole clinical picture matters more than the timing alone.
For clinicians, the practical message is about monitoring. Plan closer follow-up when baseline symptoms are more severe, and agree in advance what changes would prompt review. Earlier evidence suggests a gradual approach carries advantages: tapered discontinuation was linked to fewer somatic adverse events than abrupt discontinuation.
For patients and families, the message is not to stop medication alone. Any decision to reduce or stop should weigh the known relapse risk, treatment burdens, personal priorities and a clear follow-up plan, made with a treating psychiatrist. This paper cannot predict an individual's risk or set a taper schedule.
If you want a structured review of a long-term medication, deprescribing is a core part of my practice, and care for psychosis and schizophrenia is available in person or through online consultations. For people who need a higher level of support, inpatient psychiatric care is also available in Gurugram.
FAQ
Does rapid relapse after stopping an antipsychotic mean it is withdrawal?
Not necessarily. In this paliperidone analysis, the rapid pattern was not significantly more common among relapses after discontinuation than among relapses during continued treatment, and it tracked baseline severity. The study could not rule out withdrawal effects that PANSS does not capture.
Is it safe to stop an antipsychotic after this study?
This study does not show that. It analysed only people who relapsed and does not estimate overall relapse risk, which rises after discontinuation. Any change should be planned with a treating psychiatrist.
How fast was relapse in the rapid group?
The median time to relapse in the rapid class was 56 days for LAI paliperidone (IQR 29 to 79) and 10 days for oral paliperidone (IQR 8 to 17). The labels are specific to each cohort and should not be combined into one cutoff.
Why did baseline severity matter?
People in the rapid class began the trials with higher symptom scores and showed more fluctuation. The authors read this as underlying illness susceptibility, and note that low ratings at screening followed by regression toward usual levels may also contribute.
Does tapering matter?
Evidence from a separate individual participant meta-analysis suggests tapered discontinuation can reduce somatic adverse events. This 2026 paper did not test taper speed, so it cannot say whether tapering changes relapse trajectories.
Do these results apply to clozapine or other antipsychotics?
No. All five trials studied paliperidone, and the authors caution that the findings may not generalise to other antipsychotics or clinical settings, especially abrupt clozapine cessation, which this analysis did not examine.
What is latent class mixed modelling?
It is a data-driven method that groups people with similar patterns of symptom change over time. The researchers applied it to repeated PANSS scores to find rapid and delayed relapse trajectories, although such models need analytic choices and can overfit small samples.
What should I do if symptoms worsen after reducing medication?
Contact your treating psychiatrist promptly. If symptoms are severe or you feel unsafe, go to the nearest emergency department or call Tele-MANAS on 14416 in India. You can also book a consultation to review a medication plan.
This article is for education and is not individual medical advice. Do not stop or change psychiatric medication without guidance from your treating doctor. Source paper: Louie K, Jauhar S, Rubio J, et al. Relapse trajectories and antipsychotic discontinuation in schizophrenia: an individual participant data meta-analysis. The Lancet Psychiatry 2026;13:829-840.
