Predicting Antidepressant Treatment Response Using MRI
Author Information
Author(s): Maarten G. Poirot, Daphne E. Boucherie, Matthan W. A. Caan, Roberto Goya‐Maldonado, Vladimir Belov, Emmanuelle Corruble, Romain Colle, Baptiste Couvy‐Duchesne, Toshiharu Kamishikiryo, Hotaka Shinzato, Naho Ichikawa, Go Okada, Yasumasa Okamoto, Ben J. Harrison, Christopher G. Davey, Alec J. Jamieson, Kathryn R. Cullen, Zeynep Başgöze, Bonnie Klimes‐Dougan, Bryon A. Mueller, Francesco Benedetti, Sarah Poletti, Elisa M. T. Melloni, Christopher R. K. Ching, Ling‐Li Zeng, Joaquim Radua, Laura K. M. Han, Neda Jahanshad, Sophia I. Thomopoulos, Elena Pozzi, Dick J. Veltman, Lianne Schmaal, Paul M. Thompson, Henricus G. Ruhe, Liesbeth Reneman, Anouk Schrantee
Primary Institution: Amsterdam UMC, Department of Radiology and Nuclear Medicine, University of Amsterdam
Hypothesis
Can machine learning approaches applied to pre-treatment cortical structural MRI-derived measures predict pharmacotherapeutic treatment response better than chance?
Conclusion
Cortical structural MRI alone is not a reliable predictor of individualized pharmacotherapeutic treatment response in major depressive disorder.
Supporting Evidence
- Cortical structural MRI measures showed no predictive value for treatment response in the overall population.
- Performance predicting treatment response was not significantly better than chance across various machine learning configurations.
- Significant predictive performance was observed only in the extreme (non‐)responders subpopulation.
Takeaway
This study looked at brain scans to see if they could help predict how well people with depression would respond to antidepressants, but it found that they couldn't do this reliably.
Methodology
The study used machine learning methods to analyze cortical morphometry from MRI scans of 262 patients with major depressive disorder across six sites.
Potential Biases
Potential biases may arise from the variability in clinical assessment and treatment protocols across different sites.
Limitations
The study's findings may not generalize to all patients with MDD due to the specific subpopulations analyzed and the variability in treatment response across cohorts.
Participant Demographics
Mean age of participants was 36.5 years, with 59% being female.
Statistical Information
P-Value
0.66
Statistical Significance
p=0.66
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website