A Bayesian analysis of pharmacoeconomic data from a clinical trial on schizophrenia.

Pharmacoeconomic studies are performed in a higher frequency to assess the economic interest of new drugs. However, a standard methodology does not still completely exist. We present here the principles and results of a cost-effectiveness Bayesian analysis on data from 146 patients (interim analysis) collected during a clinical trial. This trial was originally planned to enrol 245 patients with predominantly negative schizophrenia symptoms and involved four treatment groups (a new treatment given at low dose and high dose, a comparator and a placebo). First, some prior distributions of the cost-effectiveness ratio were numerically deduced from the effectiveness parameter clinical priors (based on investigators' opinions and questionnaires before going to blind breaking) and from cost function priors. The costs taken into account were hospitalizations, sick leave days, treatments, visits to the doctor, laboratory exams and suicide attempts. The effectiveness parameter was the change from baseline on SANS (scale for the assessment of negative symptoms). Posterior distributions were elaborated for the cost-effectiveness ratio by combining the cost-effectiveness ratio priors and likelihood together using the Bayes theorem. Results lead to a conclusion in favour of the new treatment given at high dose.