Background: We sought to identify haloperidol-treated subjects who relapsed within 6 weeks of placebo replacement and those who did not, using multivariate analysis.
Methods: In the week prior to discontinuation of haloperidol treatment, global behavioral ratings and a lumbar puncture for cerebrospinal fluid monoamine metabolities were obtained in 88 patients with chronic schizophrenia. Logistic regression analyses were used to evaluate two competing models of relapse prediction. The models were then compared using receiver operating characteristic analysis and a final combined model was derived.
Results: The behavioral model was less variable in its prediction than the cerebrospinal fluid monoamine model. The final model consisted of increased psychosis, decreased anxiety, higher cerebrospinal fluid homovanillic acid levels, and lower cerebrospinal fluid 5-hydroxyindoleacetic acid levels.
Conclusions: Several monoamine systems are involved in psychotic relapse within 6 weeks of haloperidol withdrawal. Future studies of relapse prediction should include both clinical and biological measures to fully assess relapse risk.