Back to search

University of Alabama Birmingham

Statistical methods for Evaluating Confounding in Nutritional Epidemiology: Examining the Association of Dietary Measures and Incident Type 2 Diabetes

Abstract

dc:description.abstract

Nutrition is a key factor in the development of noncommunicable diseases (NCDs) such as cardiovascular disease, diabetes, obesity, and cancer, which are continuing to increase globally. Research on dietary patterns is crucial for informing what aspects of diet are most critical for preventing and possibly reversing chronic disease. Dietary patterns focus on the types of food that are consumed which reflects the complexity and cultural aspects of diet rather than focusing on a single food or nutrient. Observational cohort studies are able to observe typical dietary intake and thus better reflect the real association between diet and chronic disease. In order to elucidate the association between diet and chronic disease, observational studies can account for a range of factors influencing diet through multivariable modeling. Issues of confounding are one of the primary limitations of observational studies. Even after employing methods such as risk factor adjustment for measured confounding, addressing potential unmeasured confounding remains an issue and can be related to either imperfectly measured variables or unmeasured variables. Given the limitations that have traditionally accompanied sensitivity analysis and other methods for addressing unmeasured confounding, such as lack of comparability, iv interpretability, and potential complexity of assumptions, VanderWeele et al. proposed a new measure, the E-value, for sensitivity analysis that requires minimal assumptions to evaluate the evidence for causality. The outline of this dissertation is as follows. After review of the literature, the first paper undertakes a scoping review to systematically assess how the E-value has been implemented in nutritional epidemiology thus far. The second paper evaluates the utility of the E-value in relation to the association of dietary measures and incident type 2 diabetes by comparing known associations of each potential confounder to an E-value from the model without the confounder. The final paper implements the E-value in relation to the comparison of the association of multiple dietary measures with incident type 2 diabetes in the REasons for Geographic And Racial Differences in Stroke (REGARDS) study. In concluding, public health recommendations and suggestions for future research are provided in relation to the dissertation research.

Degree

thesis:*
Level thesis:degree_level
Dissertation

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Tison, Stephanie E
Contributors dc:contributor
  • Stacey S Cofield
  • Suzanne Judd
  • April P Carson
  • D Leann Long
  • James M Shikany

Subjects

dc:subject × 5

Identifiers

dc:identifier.*

Chain of custody

source
Harvested from
University of Alabama Birmingham
Base URL
digitalcommons.library.uab.edu/do/oai/
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Tison, Stephanie E. Statistical methods for Evaluating Confounding in Nutritional Epidemiology: Examining the Association of Dietary Measures and Incident Type 2 Diabetes. Dissertation thesis, https://digitalcommons.library.uab.edu/etd-collection/518