The Associations Between Diabetes, Cardiovascular Disease and Occupational Hearing Loss in a Cross-Sectional Analysis of Department of Energy Nuclear Workers, With a Mendelian Randomization of Publicly Available Data to Enhance Observational Findings
Date of Degree
Doctor of Philosophy (Ph.D.)
Environmental, Occupational, and Geospatial Health Sciences
Occupational Health and Industrial Hygiene | Public Health
Nuclear Workers, Hearing Loss, Hypertension
Title: The Associations between Diabetes, Cardiovascular Disease and Occupational Hearing Loss in a Cross-sectional analysis of Department of Energy Nuclear Workers, with a Mendelian Randomization of Publicly Available Data to Enhance Observational Findings
Author: Jonathan Corbin
Advisor: Ghada Soliman, MD, PhD, R.D., CDN
Introduction: Occupational Hearing Loss (OHL) is the most common workplace health condition, and the overall impact of hearing loss (HL) on health is underappreciated. Diabetes and cardiovascular disease (CVD) are two of the most common causes of death and are largely preventable and treatable. Previous research has evaluated associations between OHL and diabetes, and OHL and CVD; however, the results are inconsistent and frequently reflect study design limitations. Our research seeks to further clarify the relationships between diabetes and OHL, and CVD risk factors and OHL. Positive associations might increase intervention opportunities for health and safety practitioners and illuminate the importance of addressing work-related and general health concerns simultaneously. We introduce the use of Mendelian Randomization (MR) techniques as a novel tool to enhance the findings of observational occupational health studies.
Objectives: We sought to examine the relationship between diabetes and OHL, and CVD risk and OHL, in a group of former Department of Energy Nevada Test Site (NTS) workers participating in a medical screening program. Cross-sectional studies are at risk of unmeasured confounding and reverse causality, thus, we also conducted bi-directional Mendelian Randomization (MR) analyses to examine our findings in an unconfounded manner and test for directionality of associations.
Methods: The first study is a cross-sectional analysis of diabetes and OHL in 1,132 participants from the medical screening program for former NTS workers (Chapter 2). The second study is a cross-sectional analysis between selected cardiovascular factors or diseases (hypertension, hypercholesterolemia, and heart attack) and OHL in the same NTS data set (Chapter 3). For the first two studies, we conducted bi-variate and multi-variable logistic regressions; odds ratios (OR), confidence intervals and p-values are presented. Chapter 2 also presents risk of OHL by participant characteristics. In the third chapter, we conducted two-sample bi-directional instrumental variable analyses using genome-wide significant genetic instruments for diabetes, hypertension, high cholesterol, hearing difficulty/trouble with background noise and hearing aid use. Inverse variance weighted estimates were performed, with weighted median, MR Egger and MR Presso tests performed as sensitivity analyses.
Results: In Chapter 2, we found a positive and significant association between diabetes and OHL in the unadjusted model (OR 1.5, CI 1.19-1.95, p=.003), and this result was largely unchanged in subsequent adjusted models, while controlling for up to eight potential confounders. Age, gender, race, education, smoking status, and job category were all also significantly associated with OHL. In Chapter 3, we found a positive and significant association between self-reported hypertension and OHL across all models (Model 4: OR 1.47, CI: 1.08-1.9, p=.013, adjusted for age, gender, smoking, body mass index, education, job category and work years) and measured Stage 2 hypertension (>140 mm/hg) and OHL across all models (Model 4: OR 1.54, CI: 1.04-2.29, p=.03). We found no significant association between hypercholesterolemia and OHL and a positive but a non-significant association between self-reported heart attack and OHL. In our MR analyses, genetically predicted type 2 diabetes (T2DM), hypertension and high cholesterol were not associated with HL. Likewise, genetically predicted HL was not associated with T2DM, hypertension or cholesterol. Although MR adequately provides unconfounded estimates, selection bias cannot be ruled out from our findings.
Conclusions: Our observational studies found a clear association between diabetes and OHL, as well as between hypertension and OHL, but no association of hypercholesterolemia or self-reported heart attack with OHL. Contrary to our expectations, our MR study did not further confirm associations between diabetes and HL or hypertension and HL, although we had similar non-significant confirmatory results for hypercholesterolemia and HL. Regardless, our observational data suggests OHL is significantly elevated in those with diabetes and hypertension, even if we unsure of causality. Thus, our investigation supports new opportunities for awareness and intervention for diabetes, hypertension and HL. Furthermore, we have illustrated a significant relationship between work and non-work health risks, and findings may be of use to both occupational and primary care physicians. Finally, we illustrated the utility of using genetic data to conduct MR studies in efforts to better explain and understand findings of observational public health studies.
Corbin, Jonathan, "The Associations Between Diabetes, Cardiovascular Disease and Occupational Hearing Loss in a Cross-Sectional Analysis of Department of Energy Nuclear Workers, With a Mendelian Randomization of Publicly Available Data to Enhance Observational Findings" (2022). CUNY Academic Works.
Available for download on Friday, May 19, 2023