Observational studies commonly report a J‑ or U‑shaped association between alcohol use and symptoms of anxiety and depression, where moderate drinkers appear to have lower symptom severity than abstainers. One plausible explanation for this pattern is sick‑quitter bias: individuals may reduce or stop alcohol consumption in response to declining physical health, and that declining health itself may be associated with worse mental health outcomes. The investigators aimed to evaluate whether the observed association reflects a causal effect of alcohol use on anxiety and depression symptoms or whether residual confounding driven by sick‑quitter bias explains the pattern.
This study used a causal inference approach to examine sustained alcohol use strategies and subsequent symptom severity, explicitly assessing how baseline health status modifies the abstinent versus moderate‑risk contrast.
The analysis used data from 4,673 participants in the CHASING COVID Cohort, a U.S. longitudinal study with repeated measures collected from September 2021 through December 2023. Alcohol use was assessed at five timepoints with the AUDIT‑C instrument and categorized into four groups: abstinent, low‑risk, moderate‑risk, and high/severe‑risk drinking.
Mental health outcomes were measured at follow‑up using validated symptom scales: GAD‑7 for anxiety and PHQ‑8 for depression. The study evaluated symptom score differences associated with sustained alcohol use strategies defined over time.
The investigators emulated a target trial to estimate the effects of sustained alcohol use strategies on anxiety and depression symptom severity. To handle time‑varying confounding and informative loss to follow‑up, they applied artificial censoring combined with stabilized inverse probability weighting. This approach aims to approximate causal contrasts that would arise from assigning participants to sustained alcohol use strategies, while adjusting for measured time‑dependent covariates and censoring.
The main contrast highlighted in the analysis compared abstinent participants to those classified as moderate‑risk drinkers. The authors report estimated symptom score contrasts and 95% confidence intervals for both PHQ‑8 and GAD‑7 outcomes.
To investigate the role of sick‑quitter bias, the authors stratified analyses by baseline health status. Specifically, they compared results in participants without underlying conditions or poor self‑rated health at time zero to those with poorer baseline health. The logic is that if declining health drives cessation or reduction of drinking and independently increases symptoms, the abstinent versus moderate‑risk contrast should be larger among participants who already have poorer health at baseline and attenuated among healthier participants.
In the full sample, the analysis reproduced a J‑shaped pattern: predicted symptom scores were lowest among moderate‑risk drinkers and higher among abstinent and high‑risk groups. Reported contrasts for abstinent versus moderate‑risk were:
When restricting to participants without underlying conditions or poor self‑rated health at baseline, the abstinent versus moderate‑risk contrasts attenuated to values near zero for both outcomes, indicating minimal difference in symptom severity between abstinent and moderate‑risk drinkers in the healthier subgroup.
In contrast, among participants with poorer baseline health the J‑shaped pattern was amplified. Reported abstinent versus moderate‑risk contrasts in this group were substantially larger:
Across all analyses, high‑risk drinking was consistently associated with higher anxiety and depression symptom scores.
The attenuation of the J‑shaped pattern among participants without poorer baseline health is consistent with sick‑quitter bias contributing to the apparent protective association of moderate drinking in observational studies. The amplification of the J‑shape among participants with worse baseline health supports the interpretation that declining health leads some people to reduce or stop drinking and that this decline is independently associated with worse mental health symptoms.
The authors caution that these findings underscore the importance of accounting for health‑related selection into abstinence or low‑risk drinking when interpreting observational associations between alcohol use and mental health. They explicitly warn against interpreting moderate drinking as causally protective for symptoms of anxiety and depression based on observational data that do not adequately address declining health confounding.
As reported on the preprint server, this manuscript is a preprint and has not been peer reviewed; it should not be used to guide clinical practice without further validation.
The authors report that datasets generated and/or analyzed are available in the Zenodo repository (DOI: 10.5281/zenodo.6127734), with some data elements restricted per funder requirements and available from the authors on reasonable request. The Institutional Review Board of the City University of New York Graduate School of Public Health and Health Policy approved the work. The manuscript includes declarations that all necessary participant consent and reporting guidelines were followed.
Declared funders include the National Institute of Allergy and Infectious Diseases (UH3AI133675), National Institute of Mental Health (RF1MH132360), National Institute of Child Health and Human Development (P2C HD050924), Pfizer, CUNY Institute for Implementation Science in Population Health, CUNY Graduate School of Public Health and Health Policy, and a COVID‑19 grant program at CUNY. One author (DN) reported consulting fees from AbbVie and Gilead and an institutional research grant from Pfizer; other authors declared no conflicts. The sponsors were reported to have no role in study design, data collection, analysis, interpretation, manuscript writing, or decision to publish.