Dissertations and Theses

Date of Degree

5-30-2025

Document Type

Dissertation

Degree Name

Doctor of Philosophy (Ph.D.)

Department

Epidemiology and Biostatistics

Advisor(s)

Luisa N. Borrell, DDS, PhD

Committee Members

Heidi E. Jones, PhD, MPH

Linda Valeri, PhD

Katarzyna E. Wyka, PhD

Subject Categories

Community Health and Preventive Medicine | Epidemiology | Genetics | Medicine and Health | Public Health | Public Health Education and Promotion | Race and Ethnicity

Keywords

Telomere length, Racial/ethnic inequities, social epidemiology, mortality, cardiovascular diseases, cancer, competing risk analysis

Abstract

BACKGROUND

Telomeres, the protective non-coding DNA sequences at chromosome ends, shorten with each cell division, making telomere length (TL) a biomarker of biological aging and an indicator of health outcomes and mortality risk. Despite research linking TL to all-cause, cardiovascular disease (CVD), and cancer-specific mortality, findings remain inconsistent, with significant variability across age groups, follow-up periods, and study populations. Additionally, TL's potential role as a mediator has not been comprehensively examined concerning racial/ethnic differences and socioeconomic status (SES) indicators like education with inequities in mortality. Gaps remain in understanding how TL mediates the influence of racial/ethnic background and SES on mortality and how TL's role in CVD-specific mortality can be accurately assessed using advanced statistical methods that account for competing risks. This dissertation addresses these gaps, employing robust statistical approaches to clarify the nuanced relationships between TL, race/ethnicity, SES, and mortality outcomes, with implications for understanding health inequalities in the U.S.

METHODS

This retrospective cohort study utilized data from the National Health and Nutrition Examination Survey (NHANES) 1999–2002 linked to the 2019 Linked Mortality Files (LMF). NHANES is a nationally representative survey that collects health information from a diverse population. A probabilistic matching algorithm was used to link NHANES participants to the LMF, allowing for mortality follow-up. The study was restricted to adults aged 25 years or older.

For Aim 1, Cox proportional hazards regression was used to examine the association between TL and all-cause, CVD-specific, and cancer-specific mortality. Consistent with previous studies, TL was standardized and expressed in kilobase pairs (kbp). We tested for effect modification by age group and race/ethnicity, and reported hazard ratios (HR) and 95% confidence intervals (CI).

For Aim 2, we used Cox proportional hazards regression and causal mediation analysis to investigate the joint association of race/ethnicity and education with all-cause mortality risk and the contribution of TL to this relationship. For this analysis, we centered TL measurements to the mean and reported adjusted hazard ratios (aHR) and 95% CI. In the mediation analysis allowing for exposure-mediator interaction, we reported the total, direct, and indirect effects, measured as HR and 95% CI.

For Aim 3, we employed a multistate model to evaluate transitions from an event-free state to CVD-specific and non-CVD-specific mortality, estimating cause-specific HR, adjusted for sociodemographic and health risk factors.

RESULTS

At baseline, the average telomere length (TL) in the study population was 5.77 kbp, with longer TL observed among younger individuals (25-44 years), females, Blacks, and those with higher education. Shorter TL was associated with former smoking and chronic conditions like high cholesterol, diabetes, hypertension, and CVD. By the end of the study (December 31, 2019), one-quarter of participants had died, with one-third of deaths due to CVD and one-fifth to cancer; deceased participants had shorter TL compared with those alive at follow-up.

For Aim 1, Kaplan-Meier analyses indicated that each quartile decrease in TL correlated with higher risks of all-cause, CVD-specific, and cancer-specific mortality (P < 0.001). After adjusting for baseline sociodemographic characteristics, health risks, and chronic conditions, each kbp decrease in TL was associated with a 1.36-fold increase in the rate of all-cause mortality (HR: 1.36, 95% CI: 1.17–1.58). Adults in the shortest TL quartile (Q1) had a 1.57 times higher rate of all-cause mortality than those in the longest quartile (Q4) (HR: 1.57, 95% CI: 1.25–1.98). For cardiovascular disease (CVD)-specific mortality, each kbp decrease in TL was associated with a 1.49 times higher mortality rate (HR: 1.49, 95% CI: 1.17–1.91), and adults in the shortest TL quartile had a 1.72 times higher rate of CVD-specific mortality relative to those in the longest quartile (HR: 1.72, 95% CI: 1.18–2.50). Regarding cancer-specific mortality, adults with the shortest TL had a 1.85-fold higher rate compared with those with the longest TL (HR, 1.85; 95% CI: 1.15–2.98). No significant effect modification was observed by age or race/ethnicity.

For Aim 2, in the age- and sex-adjusted models, White participants with less than a high school education had 2.34 times higher rates of all-cause mortality (95% CI: 1.99, 2.75) compared with Whites with some college education. When compared with White adults with at least a college degree, Black adults with at least a college degree had a 1.35 higher rate of dying (95% CI: 1.00, 1.82). No significant mortality differences were observed for Mexican American adults across different educational attainment levels compared with White participants with some college education. Mediation analysis showed no significant indirect effect of TL on the joint association of race/ethnicity and education with all-cause mortality rate.

At 250 months, the cumulative incidence function (CIF) for CVD-specific mortality was 4.9% (95% CI: 3.9%-6.1%) and for non-CVD mortality was 14.4% (95% CI: 12.1%-17.2%), indicating a lower risk of CVD-specific mortality within this timeframe. Among TL quartiles, individuals in the lowest TL quartile had a 250-month cumulative incidence of 19% for CVD-specific mortality compared with 5% in the highest TL quartile (p < 0.01). In adjusted models, each unit decrease in TL increased CVD-specific mortality by 1.57 times (95% CI: 1.24-1.98) and non-CVD mortality by 1.33 times (95% CI: 1.13-1.56). For TL quartiles, those in the shortest TL quartile had a 1.88-fold higher rate of CVD-specific mortality and a 1.57-fold higher rate of non-CVD mortality than those in the longest TL quartile.

DISCUSSION

The findings of this dissertation underscore the significant role of TL in predicting mortality outcomes, notably all-cause, CVD-specific, and cancer-specific mortality. These findings are consistent with previous evidence linking telomere attrition to increased mortality risk, supporting the potential utility of TL as a biomarker for health risk assessment, especially in relation to CVD-specific mortality. The observed inverse association between TL and CVD-specific mortality in particular underscores TL’s potential as a biomarker for cardiovascular risk, suggesting that TL may offer insight into individual aging processes and vulnerability to chronic disease. The use of a multistate model, accounting for competing risks of different mortality causes, enhances the robustness of these findings and emphasizes the complex nature of these associations.

The study also provides a deeper understanding of how social determinants, specifically race/ethnicity and education, intersect with biological aging indicators like TL to shape all-cause mortality risk. The joint association of race/ethnicity and education on rate of all-cause mortality points to the critical role of social factors in health outcomes, beyond what biological measures like TL alone can explain. The lack of significant mediation of TL in the relationship between race/ethnicity, education, and all-cause mortality suggests that while TL is informative about biological aging, it does not fully capture the broader inequities rooted in social determinants. This finding highlights the importance of a comprehensive approach to health inequalities, integrating biological, social, and environmental factors.

In conclusion, this study reinforces the value of TL as a predictor of mortality, particularly in the context of CVD outcomes, while also revealing the limitations of TL as a standalone predictor of health inequalities. The complex interplay between biological aging and social determinants of health warrants further investigation to identify how factors like race/ethnicity and socioeconomic status may amplify or attenuate biological risks. Future research should continue to explore the combined effects of biological and social factors on health, with a focus on elucidating the mechanisms and pathways through which social conditions influence biological aging and mortality risk. Addressing these intersecting factors can ultimately lead to targeted public health interventions aimed at reducing health inequalities and promoting longevity across populations.

Download

Available for download on Saturday, December 19, 2026

Share

COinS