Effect of background therapy with Non-steroidal anti-inflammatory drugs (NSAIDs) and other anti-inflammatory agents on COVID-19 outcomes

Background Inflammation plays a complex, incompletely understood role in the pathogenesis of acute COVID-19 and Post-Acute Sequelae of SARS-CoV-2 infection (PASC or “Long COVID”). Systemic acute inflammation resulting in cytokine storm, hypercoagulability and endothelial damage is thought to be a central mechanism for severe morbidity and mortality in acute COVID-19. Anti-inflammatory medications taken routinely for chronic conditions prior to contracting COVID-19 (“background medications”) may modulate acute COVID-19 outcomes. Methods Using data from the National COVID Cohort Collaborative (N3C) enclave, we estimated effects of six classes of background medications on acute COVID outcomes. Medication classes included aspirin, celecoxib, other NSAIDS, steroids, immune suppressants, and antidepressants. Acute COVID outcomes included probability of hospital admission, inpatient mortality, and mortality among diagnosed COVID patients. Each medication class was compared to benzodiazepines (excluding midazolam) which served as a comparator/control. Only adult COVID patients with pre-existing osteoarthritis and without any diagnosed autoimmune disease were included in the analyses. Random effects logistic regression models were used to adjust for covariates and data contributing organization. Medication effects also were estimated for COVID-negative cases. Results Non-aspirin NSAIDS were associated with lower mortality among diagnosed COVID-19 patients: adjusted Odds Ratio (aOR)=0.32 (p=.032) for celecoxib; aOR=0.51 (p<.001) for NSAIDS other than aspirin and celecoxib. For inpatient mortality: aOR=0.34 (p=.060) for celecoxib and aOR=0.74 (p=.200) for other non-aspirin NSAIDS. Similar effects were observed for COVID-negative cases, including for inpatient mortality: aOR=0.21 (p<.001) for celecoxib and aOR=0.34 (p<.001) for other non-aspirin NSAIDS. Secondary analyses examined alternative explanations for results. Discussion Protective effects were observed for non-aspirin NSAIDS, especially celecoxib. However, those estimated effects implicitly assume the medication classes did not differ on the probability a true COVID-19 case was diagnosed. The similarity of COVID-positive and COVID-negative results suggest possible missing covariates. However, such similarity plausibly could stem from a medication having both “direct” and “indirect” effects on COVID outcomes. Adjudicating among the alternative interpretations would require data beyond those available. However, the effects observed for non-aspirin NSAIDS, while possibly biased, rationalize further investigation using study designs constructed to overcome the limitations of existing datasets.