[PMH12] Comparison of polypharmacy vs monotherapy on occurrence of relapse in schizophrenic patients: advantage of propensity scoring adjustment

[PMH12] Comparison of polypharmacy vs monotherapy on occurrence of relapse in schizophrenic patients: advantage of propensity scoring adjustment

2009 Value in health

Sarlon, E. | Millier, A. | Cristeau, O. | Toumi, M. | Volume: 12, Issue: 7, Pages: A352, Polypharmacy, Schizophrenia,

OBJECTIVES: The objective was to compare occurrence of relapse according to
antipsychotic treatment (monotherapy vs polypharmacy), in a 2 year observational
cohort of 288 French schizophrenic patients, using different methods of adjustment.
METHODS: Relapse was defined by a usual, clinically well reproducible and validated
definition. The occurrence of relapse between patients receiving one antipsychotic to
patients receiving more than one antipsychotic was compared according to a Cox
model including or not the propensity score (PS). The propensity score was calculated
on socio-economic and demographic information, clinical variables, quality of life,
medication information and attitude toward treatment. Cox models were built according
to four usual methods to calculate the propensity score: 1-stepwise PS, continuous,
2-non stepwise PS, continuous, 3-non stepwise PS, quintiles 4-non stepwise
PS quintiles used to stratify the sample. We also used standard adjustment using all
variables included in the model estimating the PS. Likelihood, Akaike’s information
criterion (AIC) criterion and Schwatz’s Bayesian Criterion (SBC) criterion were used
to compare the models. RESULTS: The standard cox model reports significant
decrease of relapse in patient receiving polytherapy. Consistently the four models
based on propensity scoring methods did not find any difference between the two
populations. The model that optimizes all criteria (2logL, AIC and SBC) is the model
using the stratification on the propensity score. CONCLUSIONS: Polytherapy is not
associated to a reduction of relapse while consistently literature reports major risk
associated to polypharmacy including increased mortality. The use of propensity
scoring is well established for observational data and this study illustrate that standard
regression could be misleading.