Predictive biomarkers of COVID-19 prognosis identified in Bangladesh patients and validated in Japanese cohorts

Abstract Despite high vaccination rates globally, countries are still grappling with new COVID infections, and patients diagnosed as mild dying at home during outpatient treatment. Hence, this study aim to identify, then validate, biomarkers that could predict if newly infected COVID-19 patients would subsequently require hospitalization or could recover safely with medication as outpatients. Serum cytokine/chemokine data from 129 COVID-19 patients within 7 days after the onset of symptoms in Bangladesh were used as training data. The majority of patients were infected with the Omicron variant and over 88% were vaccinated. Patients were divided into those with mild symptoms who recovered, and those who deteriorated to moderate or severe illness. Using the Lasso method, 15 predictive markers were identified and used to classify patients into these two groups. The biomarkers were then validated in a cohort of 194 Covid patients in Japan with a predictive accuracy that exceeded 80% for patients infected with Delta and Omicron variants, and 70% for Wuhan and Alpha variants. In an environment of widespread vaccination, these biomarkers could help medical practitioners determine if newly infected COVID-19 patients will improve and can be managed on an out-patient basis, or if they will deteriorate and require hospitalization.