Summary

Currently, treatment for MDD relies primarily on trial-and-error, with antidepressants (AD) estimated to have a response rate of 42%-53%. Combining clinical and demographic factors with feature selection techniques and machine-learning algorithms can enable us to better understand MDD multifactorial complexity. An average accuracy of 60.12% was achieved for a group of machine learning models for five antidepressants.

Discover more about PREDICTIX Genetics decision support tool. 

Abstract

The treatment of major depressive disorder (MDD) is complex and multifactorial, making it difficult to tailor medication to each patient’s needs. MDD treatment is mainly based on trial and error, with response rates of 42-53% for antidepressants. 

Objectives

We used a data-driven approach to analyze combinations of genetic, clinical, and demographic factors to develop an accurate predictor of response to a panel of antidepressants and optimize treatment selection. 

Methods

We analyzed the response patterns of patients to three antidepressant medications in the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study, and used state-of-the-art machine learning (ML) tools to develop a predictive algorithm. To validate our results, we evaluated the algorithm’s capacity to predict individualized antidepressant responses on a separate set of 530 patients in STAR*D, comprising of 271 patients in a validation set and 259 patients in the final test set. 

Results

Based on this assessment, the average balanced accuracy rate was 72.3% (SD 8.1) and 70.1% (SD 6.8) across the different medications in the validation and test set, respectively (p < 0.01 for all models). In order to validate our design scheme, we gathered data from the Pharmacogenomic Research Network Antidepressant Medication Pharmacogenomic Study (PGRN-AMPS) of patients treated with citalopram, and applied the algorithm’s citalopram model. This external validation yielded highly similar results for STAR*D and PGRN-AMPS test sets, with a balanced accuracy of 60.5% and 61.3%, respectively (both p’s < 0.01). 

Conclusion

To improve accuracy in antidepressant prescription,  these study findings support the feasibility of using ML algorithms applied to large datasets with genetic, clinical, and demographic characteristics.