Predictors of symptomatic and wellbeing remission in real-world samples of patients living with schizophrenia treated with aripiprazole once-monthly by means of constrained confidence partitioning

Abstract

Introduction

An important goal in the treatment of patients living with schizophrenia is to achieve remission of schizophrenia symptoms and patient well-being. The possibility to predict which patients will achieve remission may inform treatment decisions. We derived different models for this on the basis of data from a German real-world study, part of which were tested for their predictive power by independent data obtained from a similar Canadian study.

Methods

Here, we used a sample of patients living with schizophrenia who participated in a 6-month non-interventional study of aripiprazole once-monthly. Data of patients with complete datasets (n = 194) were used to predict remission of symptoms (cross-sectional Andreasen criteria as measured by the Brief Psychiatric Rating Scale, BPRS) and well-being (as measured by the WHO-5 well-being index) at 6 months using logistic regression as well as the constrained confidence partitioning (c²p) method.

Results

Logistic regression yielded a model with the variables remission stats at baseline, Global Assessment of Functioning (GAF) score at baseline and age. Quality indicators suggested acceptable model quality, and we were able to validate the model using external data from the Canadian study. Models for remission of well-being and combined remission of symptoms and well-being were of poor quality. Modeling using c²p yielded defined groups of patients with differential likelihoods of achieving remission of symptoms, well-being, or both. In general, patients with lower scores in core symptoms, higher GAF scores, younger age, and higher WHO-5 scores showed increased likelihoods of achieving remission. The c²p model of symptomatic remission was also validated using external data from the Canadian study, and we demonstrated the possibility to generate and test hypotheses based on the model.

Conclusion

Despite only using a comparatively small sample, both logistic regression and c²p can produce reliable results with mathematically desirable properties. Patients with less severe core symptoms, better functioning, younger age, and better well-being may achieve symptomatic remission and remission of well-being more easily than other patients. The latter may need additional interventions in order to achieve remission.