Precision Symptom Phenotyping Identifies Early Clinical and Proteomic Predictors of Distinct COVID-19 Sequelae

Abstract

Abstract Background Post-COVID conditions (PCC) are difficult to characterize, diagnose, predict, and treat due to overlapping symptoms and poorly understood pathology. Identifying inflammatory profiles may improve clinical prognostication and trial endpoints. Methods This analysis included 1988 SARS-CoV-2 positive U.S. Military Health System beneficiaries who had quantitative post–COVID symptom scores. Among participants who reported moderate-to-severe symptoms on surveys collected 6 months post-SARS-CoV-2 infection, principal component analysis followed by k-means clustering identified distinct clusters of symptoms. Results Three symptom-based clusters were identified: a sensory cluster (loss of smell and/or taste), a fatigue/difficulty thinking cluster, and a difficulty breathing/exercise intolerance cluster. Individuals within the sensory cluster were all outpatients during their initial COVID-19 presentation. The difficulty breathing cluster had a higher likelihood of obesity and COVID-19 hospitalization than those with no/mild symptoms at 6 months post-infection. Multinomial regression linked early post-infection D-dimer and IL-1RA elevation to fatigue/difficulty thinking and elevated ICAM-1 concentrations to sensory symptoms. Conclusions We identified three distinct symptom-based PCC phenotypes with specific clinical risk factors and early post-infection inflammatory predictors. With further validation and characterization, this framework may allow more precise classification of PCC cases and potentially improve the diagnosis, prognostication, and treatment of PCC.