Rapid deterioration after antipsychotic discontinuation is often attributed to a pharmacological withdrawal phenomenon — supersensitivity psychosis — rather than to the illness returning. That distinction matters, because if it is a drug effect then slow tapering should prevent it, and if it is not then tapering speed is beside the point.
This meta-analysis used individual participant data from five randomised, double-blind, placebo-controlled discontinuation trials of oral and long-acting injectable paliperidone, obtained through the Yale Open Data Access project: 271 participants from the injectable trials and 146 from the oral trials, all stabilised on antipsychotics for more than three months before randomisation. Latent class mixed modelling on longitudinal PANSS data identified trajectories of symptom change preceding relapse.
Two classes emerged, rapid and delayed onset, with rapid relapse marked by more severe symptoms at the point of relapse. Crucially, the proportion with rapid relapse did not differ between the discontinuation and continuation groups — 20% versus 11% in the injectable trials (p=0.12) and 27% versus 26% in the oral trials (p=0.95). Symptom profiles at relapse did not differ by discontinuation status either. What did predict rapid relapse was higher baseline PANSS score (p<0.001).
If rapid relapse were a withdrawal effect, it should be concentrated among those who stopped. It was not. So the clinically useful predictor is not how the drug was stopped but how severe the illness was to begin with, and de-prescribing should be risk-stratified on baseline severity rather than paced by a universal taper.
- Stratify de-prescribing decisions by baseline symptom severity rather than by tapering schedule alone
- Do not reassure a patient that a slow taper will prevent rapid relapse — the evidence does not support a withdrawal mechanism
- Increase monitoring frequency during and after discontinuation in patients with higher baseline symptom burden
- Note that rapid relapse occurred in a quarter of those who relapsed even while continuing treatment
- This covers paliperidone only; other antipsychotics with different pharmacokinetics were not studied
The statistics, in plain English
The key comparison, 20% versus 11% with p=0.12 in the injectable trials, is a null result with a caveat: 74 people in the continuation group is a small denominator, so the trial cannot exclude a real difference. The oral comparison, 27% versus 26% with p=0.95, is much cleaner. Latent class mixed modelling identifies subgroups from the data rather than from prespecified definitions, which is powerful but can find classes that do not exist; the authors' own caution about a measurement artifact — that severity may be minimised at screening to meet trial entry thresholds, so patients with apparently low baseline scores are in fact sicker — is a real alternative explanation for the baseline severity finding.
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.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for psychiatry, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free