Rapid relapse after stopping paliperidone was not more common than during continued treatment

Design

Individual participant data meta-analysis of 5 randomised discontinuation trials; latent class modelling

Population

417 people with schizophrenia or schizoaffective disorder who relapsed (271 LAI, 146 oral paliperidone)

Primary outcome

Relapse trajectory class by discontinuation status

Effect

Rapid relapse 20% vs 11% (LAI, p=0.12); 27% vs 26% (oral, p=0.95)

This individual participant data analysis pooled five double-blind, placebo-controlled paliperidone discontinuation trials (oral and long-acting injectable) from the Yale Open Data Access database. It modelled symptom trajectories in 417 people whose schizophrenia or schizoaffective disorder relapsed.

Two patterns emerged: rapid and delayed relapse. Rapid relapse was not over-represented in those switched to placebo: 20% vs 11% in the injectable trials (p=0.12) and 27% vs 26% in the oral trials (p=0.95). Symptom profiles at relapse did not differ by discontinuation status. People with rapid relapse had higher baseline PANSS scores.

A common argument is that quick relapse after stopping an antipsychotic is a withdrawal or supersensitivity effect rather than the illness returning. These data do not fit that pattern for paliperidone.

For deprescribing, the practical point is that baseline severity, not the act of stopping, predicted who relapsed fast. That supports closer monitoring for patients who were more unwell before stabilisation.

• Before reducing an antipsychotic, note how severe the illness was before stabilisation.
• Arrange closer early review for those with high baseline severity.
• Agree early warning signs and a restart plan with the patient and family.
• Do not assume the findings extend to clozapine or other agents with different pharmacology.

Why it matters

It weakens the idea that rapid relapse after stopping paliperidone is mostly a pharmacological rebound.

Don't overread it

All five trials tested paliperidone and the comparisons were underpowered, so absence of a difference is not proof that rebound never occurs.

The statistics, in plain English

In the injectable trials, 20% vs 11% looks like a difference, but with only 74 people in the continuation group the p-value of 0.12 means chance cannot be excluded. A non-significant result in a small group is an absence of evidence, not evidence of absence.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.