Inflammatory Biomarkers as Mediators of the Effect Between High‐Dose Corticosteroid Therapy and Mortality in COVID‐19‐Related ARDS: A Causal Mediation Analysis

ABSTRACT Background Treatment guidelines for acute respiratory distress syndrome (ARDS) recommend the use of high‐dose corticosteroids based on their anti‐inflammatory effects, yet the efficacy of this therapeutic strategy in patients with COVID‐19 related ARDS is inconclusive. To more thoroughly understand the mechanism of action for high‐dose corticosteroids in patients with COVID‐19‐related ARDS, we hypothesized that the reduction of inflammatory markers induced by high‐dose corticosteroids would reduce mortality (i.e., inflammation as a mediator). Design and Methods In this retrospective cohort study, we included patients with COVID‐19‐related ARDS admitted to the intensive care unit (ICU) of an academic hospital in Rotterdam, the Netherlands between March 2020 and June 2022. High‐dose corticosteroids were defined as > 6 mg of dexamethasone/day or equivalent. Biomarkers included C‐reactive protein (CRP), d ‐dimer, ferritin, leukocyte count, interleukin‐6 (IL‐6), lactate dehydrogenase, and neutrophil‐to‐lymphocyte ratio (NLR). Using a causal mediation framework, we estimated the average mediation effect (AME) and % mediation for each biomarker, allowing to quantify the degree to which continuous levels of inflammatory markers mediate the association between receiving high‐dose corticosteroids and mortality. The models used in this framework accounted for time‐varying treatment/mediation with inverse probability of treatment weights, determined from the covariates PaO 2 /FiO 2 ratio, sex at birth, age, NLR, and ICU length of stay. Results Of the 327 patients included, 122 (37.3%) received high‐dose corticosteroids. The risk of death was higher in patients who did vs. did not receive high‐dose corticosteroids [incidence rate = 0.54, 95% confidence interval (CI) = 0.42–0.71 and 0.21, 95% CI = 0.15–0.29/person‐month, respectively; p < 0.001]. This effect remained in most analyses accounting for time‐varying treatment/mediation. The association between high‐dose corticosteroids and mortality was reduced (i.e., mediated) with lower levels of CRP (AME = −0.006, 95% CI = −0.011, −0.002; %‐mediation = −82.0%), d ‐dimer (AME = −0.002, 95% CI = −0.005, −0.001; %‐mediation = −33.1%), and IL‐6 (AME = −0.002, 95% CI = −0.004, −0.001; %‐mediation = −25.5%). There was no evidence of mediation for other biomarkers. Conclusion CRP, d ‐dimer, and IL‐6 mediated the association between high‐dose corticosteroids and mortality. When inflammation was reduced, the deleterious effect of high‐dose corticosteroids was eliminated.