Eng, Winston
(2018)
Exploring the genetic characteristics underlying a multidimensional latent chemotherapy symptom burden.
Master's Thesis, University of Pittsburgh.
(Unpublished)
Abstract
The incidence rate of cancer is expected to increase within the coming decades. While related mortality is expected to decrease with improving treatments, oncology patients are still expected to experience harmful physiological and psychological symptoms. This “symptom burden” has been shown to permanently extend into the patients’ lives, and researchers have hypothesized that a genetic component may dictate the severity of its presentation. From a public health perspective, sustaining an expanding concentration of those considered “symptom burdened” will create unsustainable stress on the current healthcare system. Hoping to describe the heterogeneity in this oncology experience, researchers have relied on statistical clustering to generate patient subgroups differing in quality-of-life. This study’s objective is to assess if there exists any association between Single Nucleotide Polymorphisms (SNP) and oncology “symptom burden” following subcategorization; more specifically, it aims to compare analyses of the latent class phenotypes using the “default” method of Multinomial Logistic Regression with those of the “novel” method of Dirichlet Regression.
A four category latent class was generated while adjusting for site from symptom clusters measured on 2111 subjects from sites UCSF_total and TOR_1 . Genotyping occurred using two versions of the Illumina exome chip: HumanCoreExome-24v1-0 (Group A) and HumanExome-12v1-1 A (Group B). Group A had 944 UCSF_total individuals, while Group B had 669 UCSF_total and 498 TOR_1 subjects. Following quality control, there were 1272 and 415 for the UCSF_total and TOR_1 cohorts respectively. Covariates included within the regression models were total number of comorbidities, Karnofsky Performance Status (KPS), sex, and the first four principal components for population substructure.
After applying both the Multinomial Logistic Regression and Dirichlet Regression approaches, neither method demonstrated statistically significant genetic association (P < 5 × 10 −8 ). However, issues concerning SNPs with very low minor allele frequencies appeared to plague both approaches, indicating that methodological corrections may be necessary in future studies. Additionally, covariate selection, sample size, and imputation may be areas of future inquiry with regards to rectifying some of the issues presented. Future aims should involve simulation and power calculations to determine how appropriate these proposed methods are for assessing genetic association in relation to oncology “symptom burden.”
Share
| Citation/Export: |
|
| Social Networking: |
|
Details
| Item Type: |
University of Pittsburgh ETD
|
| Status: |
Unpublished |
| Creators/Authors: |
|
| ETD Committee: |
|
| Date: |
28 June 2018 |
| Date Type: |
Publication |
| Defense Date: |
13 April 2018 |
| Approval Date: |
28 June 2018 |
| Submission Date: |
6 April 2018 |
| Access Restriction: |
5 year -- Restrict access to University of Pittsburgh for a period of 5 years. |
| Number of Pages: |
64 |
| Institution: |
University of Pittsburgh |
| Schools and Programs: |
School of Public Health > Biostatistics |
| Degree: |
MS - Master of Science |
| Thesis Type: |
Master's Thesis |
| Refereed: |
Yes |
| Uncontrolled Keywords: |
GWAS, symptom burden, latent class analysis, dirichlet |
| Date Deposited: |
28 Jun 2018 20:05 |
| Last Modified: |
01 May 2023 05:15 |
| URI: |
http://d-scholarship.pitt.edu/id/eprint/34168 |
Metrics
Monthly Views for the past 3 years
Plum Analytics
Actions (login required)
 |
View Item |
![[feed]](/images/feed-icon-32x32.png){kind=icon}