Assessing efficacy of chronic therapy when disease is characterized by frequent relapses: the example of schizophrenia.

In assessing the efficacy of therapy for schizophrenia, the effect of medication on relapse needs to be distinguished from the influence of relapse on medication. Typically, effective medication prevents relapse, but relapse generally induces medication. Conventional analyses using either treatment or disease as outcome do not separate these two effects. We propose an alternate approach that uses both treatment and schizophrenic relapse as random variables. Data from 58 schizophrenic patients, with up to 60 consecutive monthly determinations of antipsychotic medication and schizophrenic events, were analyzed using a bivariate transition model with random effects. This analysis revealed that the risk of current schizophrenic relapse is reduced by continuous medication (previous month and current month) but not necessarily by discontinuous medication.