Adaptation of Mixture Cure Model in Estimating Incidence and Latency for Major Depressive Disorder Público

Huo, Xingyue (Spring 2019)

Permanent URL: https://etd.library.emory.edu/concern/etds/cn69m5248?locale=es
Published

Abstract

Background: Major Depressive Disorder is a serious mental health disease that may influence people's daily lives. Cognitive behavioral therapy and medications are the main treatments for depression and have been shown to be equally effective. However, there are still open questions as to compare time to response across different treatments, and possibly identifying when to discontinue the current treatment. We sought to determine if a mixture cure model that included nonresponse might fit the data in depression treatment and examine the data to see if there are indicators of when to discontinue or switch the current treatments.

Methods: We performed Cox proportional hazard regression and conditional survival analysis on a sample of 316 patients with the major depressive disorder to estimate the incidence and latency separately. We also proposed a mixture cure model for estimating the prevalence of response and the mean time to response, simultaneously. Due to a large response rate, we were forced to use two "nonresponse filters" in order to isolate the nonresponse signal. 

Results: The results showed no clear evidence of a time to switch treatments since the survival curve were continuous and did not have any large plateaus. By comparing the results from Cox regression and conditional survival analysis, it might suggest that the incidence was responsible for the significant difference across treatments, and the CBT group was more likely to be nonresponders to traditional treatment. Without using "filters" to separate data in order to fit the mixture cure model, we cannot determine a significant association between treatment groups due to the high prevalence of response and therefore a solution closed to the boundary.

Discussion: We found it was inappropriate to use the mixture cure model with current clinical data due to high response prevalence. It emphasized the adaptation of a mixture model, the model might not fit the data with a high prevalence of the "cure". Our results also suggested the majority of patients had a response in the treatment of depression. CBT might take a longer time to respond than medications, but evidence of nonresponse in this group is limited. In addition, we did not identify when to discontinue or switch the current treatment because of the lack of a placebo group.

Table of Contents

Introduction ----1

Methods ---- 6

Results ---- 11

Discussion ---- 14

Reference ----16

Table and Graphs ----21

About this Master's Thesis

Rights statement
  • Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.
School
Department
Subfield / Discipline
Degree
Submission
Language
  • English
Research Field
Palabra Clave
Committee Chair / Thesis Advisor
Committee Members
Última modificación

Primary PDF

Supplemental Files