New insight into the CATIE study by constrained confidence partitioning. An innovative technique towards personalized antipsychotic drug therapy in schizophrenia treatment

The CATIE schizophrenia trial was a very influential randomized controlled trial in patients with chronic schizophrenia. Patients were followed for up to 18 months under treatment with a randomly assigned antipsychotic. The primary endpoint, time to discontinuation of treatment for any reason, is influenced by individual patient characteristics, external factors as well as effects of drug treatment. New insight is obtained by applying an innovative survival analysis based on constrained confidence partitioning (SA-C²P). Through this data-driven approach we identify homogeneous collectives of patients with similar patient characteristics differing from the study population in the primary endpoint, enabling us to predict patient individual outcome more precisely. A subgroup of patients treated with olanzapine featuring neither an anxiety disorder in the past month, drug abuse in the past five years nor hospitalizations in the past year discontinued drug therapy substantially later compared to patients meeting at least one of the named parameters. Moreover, differences in the primary outcome between second-generation antipsychotics increased compared to the original CATIE analysis when looking into this subgroup in the entire study sample. Our findings suggest that SA-C²P may assist in identifying relevant responder subgroups, probably missed by conventional statistical methods, making it a potential tool for personalized medicine.