Date of Award

7-5-2023

Author's School

Graduate School of Arts and Sciences

Author's Department

Psychology

Degree Name

Doctor of Philosophy (PhD)

Degree Type

Dissertation

Abstract

Obsessive-Compulsive Personality Disorder (OCPD) is the most common personality disorder, yet much remains unknown about its etiology. Although neural contributions to many other psychiatric disorders have been extensively studied, few existing studies have examined neural correlates of OCPD. Furthermore, all have had insufficient sample sizes to produce reliable results. Large samples are needed to reliably detect the expected small brain-behavior relationships. However, large neuroimaging studies often do not assess for personality disorders, although many assess for normative personality. The present study employed a Five-Factor Model of personality disorders, which conceptualizes personality disorders as maladaptive extremes of normative personality traits, to predict OCPD scores from normative personality data using machine learning techniques in a large community-based sample (n=1,606). This trained ML model was then applied to a separate dataset with normative personality and neuroimaging data (n=1,253) to generate predicted OCPD scores and subsequently examine brain structure correlates of OCPD traits. Despite a moderate ability to predict OCPD traits using normative personality data that generalizes across samples, we found limited evidence that predicted OCPD scores are associated with individual differences in brain structure. Indeed, there was only one significant univariate association wherein thicker right superior frontal cortex was associated with higher OCPD scores. Adopting ML models to generate multivariate models of brain structure resulted in imprecise models and thus no reliable associations. Collectively, these data suggest that OCPD symptoms may be predicted using normative personality data, but that OCPD personality traits may not be strongly associated with brain structure and may require exceptionally large samples to reliably identify these modest associations. Broadly, this approach exemplifies how deeply phenotyped small samples may be used to inform large national samples that may not have assessed specific phenotypes.

Language

English (en)

Chair and Committee

Ryan Bogdan

Available for download on Wednesday, June 26, 2024

Share

COinS