Predicting unsuccessful clozapine treatment after first use in adult patients with psychotic disorders

Marieke M. Beex-Oosterhuis, E. R. Rob Heerdink, Arthur R. van Gool, Rob J. van Marum

Research output: Contribution to journalArticleAcademicpeer-review


Purpose/Background Cessation of clozapine therapy and insufficient response may result in relapse of psychotic symptoms and in clinical admissions. However, discontinuation rates are high. Identifying patients at risk for unsuccessful clozapine use might enable clinicians to direct specific attention to them. Methods/Procedures Routinely collected data from a large insurance company were used to develop a simple prediction model for unsuccessful clozapine treatment in psychiatric patients 1 year after clozapine was first dispensed by a community pharmacy in the Netherlands. Multivariate logistic regression analyses were performed with the Nagelkerke R2 statistic as a measure of the predictive value of the model. Findings/Results A total of 937 patients were dispensed clozapine for the first time by their community pharmacy between January 1, 2011, and December 31, 2015 (index date). Of these, 741 patients had started their clozapine treatment in hospital before the index date (inpatient starters); the remaining 196 patients started clozapine as outpatients on the index date (outpatient starters). In 191 patients (20.4%), clozapine treatment was unsuccessful 1 year after the index date. Unsuccessful treatment was more common among outpatient starters than among inpatient starters (32.1% vs 17.3%). Using backward selection of the variables, a model consisting of 61 variables had the best predictive value overall (Nagelkerke R2 = 0.301), whereas a model consisting of 52 variables had the best predictive value in outpatient starters (Nagelkerke R2 = 0.676). Implications/Conclusions The likelihood of unsuccessful clozapine treatment after 1 year was higher among patients who started clozapine as outpatients. Despite the use of a diversity of variables and different statistical approaches, it was not possible to make a simple prediction model for unsuccessful clozapine treatment using relatively easily accessible data.
Original languageEnglish
Pages (from-to)604-608
JournalJournal of Clinical Psychopharmacology
Issue number6
Publication statusPublished - 2018

Cite this