Latent Variable Approaches for Understanding Heterogeneity in Depression: A Dissertation

Abstract

Background: Major depression is one of the most prevalent, disabling, and costly illnesses worldwide. Despite a 400% increase in antidepressant medication use since 1988, fewer than half of treated depression patients experience a clinically meaningful reduction in symptoms and uncertainty exists regarding how to successfully obtain symptom remission. Identifying homogenous subgroups based on clinically observable characteristics could improve the ability to efficiently predict who will benefit from which treatments. Methods: Latent class analysis and latent transition analysis (LTA) were applied to data from the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study to explore how to efficiently identify subgroups comprised of the multiple dimensions of depression and examine changes in subgroup membership during treatment. The specific aims of this dissertation were to: 1) evaluate latent depression subgroups for men and women prior to antidepressant treatment; 2) examine transitions in these subgroups over 12 weeks of citalopram treatment; and 3) examine differences in functional impairment between women’s depression subgroups throughout treatment. Results: Four subgroups of depression were identified for men and women throughout this work. Men’s subgroups were distinguished by depression severity and psychomotor agitation and retardation. Severity, appetite changes, insomnia, and psychomotor disturbances characterized women’s subgroups. Psychiatric comorbidities, especially anxiety disorders, were related to increased odds of membership in baseline moderate and severe depression subgroups for men and women. After 12 weeks of citalopram treatment, depression severity and psychomotor agitation were related to men’s chances of improving. Severity and appetite changes were related to women’s likelihood of improving during treatment. When functional impairment was incorporated in LTA models for women, baseline functional impairment levels were related to both depression subgroups at baseline and chances of moving to a different depression subgroup after treatment. Conclusion: Depression severity, psychomotor disturbances, appetite changes, and insomnia distinguished depression subgroups in STAR*D. Gender, functional impairment, comorbid psychiatric disorders, and likelihood of transitioning to subgroups characterized by symptom improvement differed between these subgroups. The results of this work highlight how relying solely on summary symptom rating scale scores during treatment obscures changes in depression that might be informative for improving treatment response.