Each analyte had a separate mixed effect model, and all the models were adjusted for age, sex, BMI, hypertension, diabetes, trial groups, and COVID-19 waves. Trial groups were not adjusted at screening, since participants were transfused after screening. Among the covariates, only BMI has missing data, and the prevalence of missingness was 5%. Available case method was used to handle missing data. Significance after adjusting for multiple comparison using Bonferroni correction was defined by a p value cutoff of 0·0024. IFN=interferon. IL=interleukin. IP-10=inducible protein-10. MCP-1=monocyte chemoattractant protein-1. MIP-1α=macrophage-inflammatory protein-1α TNF-α=tumour necrosis factor-α. VCAM-1=vascular cell adhesion molecule-1. VEGF-A=vascular endothelial growth factor-A.