Dissertations and Theses

Date of Degree

9-1-2026

Document Type

Dissertation

Degree Name

Doctor of Philosophy (Ph.D.)

Department

Epidemiology and Biostatistics

Advisor(s)

Zachary Shahn

Committee Members

Denis Nash

McKaylee Robertson

Angela Parcesepe

Subject Categories

Epidemiology | Population Health | Public Health

Keywords

Alcohol, depression, anxiety, causal inference, physical activity

Abstract

BACKGROUND: Anxiety and depression are among the most prevalent and disabling conditions in the United States, underscoring the importance of identifying modifiable contributors to their symptoms. While high levels of alcohol consumption have a well-established bidirectional relationship with anxiety and depression, associations at moderate levels of alcohol consumption remain controversial, with many observational studies reporting a J-shaped association in which moderate drinkers have lower risk of anxiety and depression than abstainers and heavy drinkers. However, this pattern may reflect methodological bias rather than a true protective effect. Studies of alcohol and mental health are particularly vulnerable to health-related selection into abstinence (sick-quitter bias), whereby individuals reduce or stop drinking in response to declining health that is independently associated with poorer mental health. Furthermore, the dynamic nature of alcohol consumption can result in time-varying treatment-confounder feedback, complicating conventional longitudinal analyses. Beyond these challenges, physical activity has been hypothesized to mitigate adverse mental health effects associated with alcohol use, but little is known about whether physical activity modifies associations between alcohol and symptoms of anxiety and depression. This dissertation addresses these gaps with implications for improving causal inference in studies of alcohol and mental health and for informing interventions aimed at reducing anxiety and depression symptom burden.

METHODS: This dissertation leveraged longitudinal data from the CHASING COVID Cohort Study, which followed a diverse sample of more than 5,700 U.S. adults from 2020-2023, with approximately quarterly follow-up assessments. First, associations between alcohol use and symptoms of anxiety and depression were examined according to physical activity level to assess potential effect modification (Aim 1). Second, inverse probability weighting and artificial censoring was applied to evaluate associations between alcohol use and anxiety/depression symptoms while addressing time-varying confounding and sick-quitter bias (Aim 2). Third, the parametric g-formula was used to estimate the population-level impact of sustained reductions in alcohol consumption from moderate/high risk to low-risk levels on anxiety/depression symptom burden (Aim 3).

RESULTS: For Aim 1, no evidence of interaction between alcohol use and physical activity was observed in the full sample. However, among participants without underlying health conditions, who are less susceptible to sick-quitter bias, sedentary participants with moderate-risk alcohol use had higher prevalence of anxiety/depression symptoms compared with physically active participants with low-risk alcohol use (aRR = 1.92, 95% CI: 1.50–2.45), with evidence of both additive interaction (RERI = 0.61, p< 0.01) and multiplicative interaction (interaction aRR = 1.49, p< 0.01). For Aim 2, a J-shaped association between alcohol use and anxiety/depression symptoms was observed in the full sample, with moderate-risk drinkers having the lowest predicted symptom scores. For depression (PHQ-8), the additive abstinent-to-moderate contrast was 1.38 (95% CI: 0.65, 2.14), and for anxiety (GAD-7), 0.95 (95% CI: 0.24, 1.65). However, this pattern greatly attenuated among participants without underlying conditions or poor self-rated health, with contrasts near zero for both outcomes, and was amplified among participants with poorer baseline health (PHQ-8 contrast: 3.31 [2.14, 4.32]; GAD-7 contrast: 2.83 [1.61, 3.86]). High-risk drinking was consistently associated with higher symptom burden across analyses. For Aim 3, sustained reductions in alcohol consumption from moderate/high-risk to low-risk levels were not associated with changes in population-level anxiety or depression symptom burden. Under the natural course, the estimated prevalence of moderate-to-severe depressive symptoms was 12.67% (95% CI: 12.17, 13.23), compared with 12.77% (95% CI: 12.26, 13.29) under the intervention (risk difference [RD]: 0.10 percentage points; 95% CI: −0.22, 0.50). For anxiety symptoms, prevalence was 10.40% (95% CI: 8.89, 11.89) under the natural course and 10.25% (95% CI: 8.76, 11.98) under the intervention (RD: −0.15 percentage points; 95% CI: −0.68, 0.35).

DISCUSSION: Findings from this dissertation suggest that the observed J-shaped association between alcohol use and symptoms of anxiety and depression likely reflects methodological bias due to confounding driven by health-related selection into abstinence, rather than a protective effect of moderate drinking. Associations between alcohol use and symptoms of anxiety and depression also varied according to physical activity level, with evidence of synergistic effects between sedentary behavior and moderate-risk drinking among healthier adults less susceptible to sick-quitter bias, suggesting that screening approaches jointly considering alcohol use and physical activity may improve identification of individuals at increased risk of anxiety and depression. However, hypothetical population-wide reductions in alcohol consumption were not associated with meaningful reductions in anxiety or depression symptom burden. Together, these findings highlight the importance of addressing confounding due to declining health in observational alcohol research, caution against interpreting moderate drinking as beneficial for anxiety and depression symptoms, and suggest that reducing alcohol use alone may be insufficient to substantially reduce the population burden of anxiety and depression symptoms.

Download

Available for download on Saturday, July 31, 2027

Share

COinS