April 18, 2023
1 minute read
Source/Disclosure
Curtis J, et al. Optimizing Precision Medicine in Depression Treatment: A Machine Learning Approach. Presented at: American Anxiety Society; April 13-16, 2023. washington dc
Disclosure:
Curtiss does not report related financial disclosures.
Important points:
- Ensemble machine learning models may predict second-line treatment for depression initially unresponsive to antidepressants.
- Predictive success depends on the type of treatment.
WASHINGTON — Ensemble machine learning models may indicate second-line treatment for people whose depression does not benefit from antidepressant pharmacotherapy, according to a poster presented here.
The poster was awarded the Best Poster Award from Early Career Experts by the American Association for Anxiety and Depression.
Ensemble machine learning algorithms may help identify possible treatments for people whose depression has not responded to antidepressants. Image: Adobe Stock
Dr. Joshua Curtis A postdoctoral fellow in psychology at the Depression Clinical & Research Program at Massachusetts General Hospital in Boston and colleagues evaluated an ensemble machine learning model using data from the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study. Specifically, he used data from 1,439 of her STAR*D participants who were randomly assigned to seven treatments after antidepressant treatment.
To evaluate an ensemble model that implemented multiple types of machine learning algorithms, researchers used 155 predictor variables drawn from clinical and demographic variables to identify likelihood of remission . They also evaluated the model using the top 10 predictors of remission.
“Including all the features, all the variables in the world is not necessarily the best step. said in “Using a simple t-test filter with the top 10 predictors is one of the simplest things you can do. [predictor]”
An analysis of machine learning predictions revealed the least success with extended dosing options (the area under the curve). [AUC] = 0.53-0.56) and most successful in cognitive therapy (AUC = 0.82).
“It was a good first step, but the overall predictive performance was modest,” said Curtiss. “This is consistent with other machine learning sources and research looking at this sort of thing.”
Going forward, the researchers say, future studies should identify better strategies and better predictor sets for analysis.
