Background: Epigenetic modifications may contribute to inter-individual variation that is unexplainable by presently known risk factors for COVID-19 severity (e.g., age, excess weight, or other health conditions). Estimates of youth capital (YC) reflect the difference between an individual's epigenetic - or biological - age and chronological age, and may quantify abnormal aging due to lifestyle or other environmental exposures, providing insights that could inform risk-stratification for severe COVID-19 outcomes. This study aims to thereby a) assess the association between YC and epigenetic signatures of lifestyle exposures with COVID-19 severity, and b) to assess whether the inclusion of these signatures in addition to a signature of COVID-19 severity (EPICOVID) improved the prediction of COVID-19 severity.
Methods: This study uses data from two publicly-available studies accessed via the Gene Expression Omnibus (GEO) platform (accession references: GSE168739 and GSE174818). The GSE168739 is a retrospective, cross-sectional study of 407 individuals with confirmed COVID-19 across 14 hospitals in Spain, while the GSE174818 sample is a single-center observational study of individuals admitted to the hospital for COVID-19 symptoms (n = 102). YC was estimated using the (a) Gonseth-Nusslé, (b) Horvath, (c) Hannum, and (d) PhenoAge estimates of epigenetic age. Study-specific definitions of COVID-19 severity were used, including hospitalization status (yes/no) (GSE168739) or vital status at the end of follow-up (alive/dead) (GSE174818). Logistic regression models were used to assess the association between YC, lifestyle exposures, and COVID-19 severity.
Results: Higher YC as estimated using the Gonseth-Nusslé, Hannum and PhenoAge measures was associated with reduced odds of severe symptoms (OR = 0.95, 95% CI = 0.91-1.00; OR = 0.81, 95% CI = 0.75 - 0.86; and OR = 0.85, 95% CI = 0.81-0.88, respectively) (adjusting for chronological age and sex). In contrast, a one-unit increase in the epigenetic signature for alcohol consumption was associated with 13% increased odds of severe symptoms (OR = 1.13, 95% CI = 1.05-1.23). Compared to the model including only age, sex and the EPICOVID signature, the additional inclusion of PhenoAge and the epigenetic signature for alcohol consumption improved the prediction of COVID-19 severity (AUC = 0.94, 95% CI = 0.91-0.96 versus AUC = 0.95, 95% CI = 0.93-0.97; p = 0.01). In the GSE174818 sample, only PhenoAge was associated with COVID-related mortality (OR = 0.93, 95% CI = 0.87-1.00) (adjusting for age, sex, BMI and Charlson comorbidity index).
Conclusions: Epigenetic age is a potentially useful tool in primary prevention, particularly as an incentive towards lifestyle changes that target reducing the risk of severe COVID-19 symptoms. However, additional research is needed to establish potential causal pathways and the directionality of this effect.