ABSTRACT: Over the last three years, numerous groups have reported on different predictive models of disease severity in COVID-19 patients. However, almost all such models, which relied on serum biomarkers, clinical data or a combination of both, were subsequently deemed as cumbersome, inadequate and/or subject to bias. Moreover, although serum metabolomics profiling has shown significant differences among patients with different degrees of disease severity, the use of serum metabolomics profiling to identify prognostic biomarkers has, so far, been neglected. Herein, we sought to develop highly predictive models of disease severity by integrating routine laboratory findings and serum metabolomics profiling which identified several metabolites including K_4_aminophenol, acetaminophen and cytosine as potential biomarkers of disease severity in COVID-19 patients. Two models were subsequently developed and internally validated on the basis of ROC-AUC values. The predictive accuracy of the first model was 0.998 (95% CI: 0.992 to 1.000) with an optimal cut-off risk score of 4 biomarkers from among 8 linearly-related biomarkers (D-dimer, ferritin, neutrophil counts, Hp, sTfR, K_4_aminophenol, acetaminophen and cytosine). The predictive accuracy of the second model was 0.996 (95% CI: 0.989 to 1.000) with an optimal cut-off risk score of 3 biomarkers from among 6 biomarkers (D-dimer, ferritin, neutrophil counts, Hp, sTfR and cytosine). The two models are of high predictive power, need a small number of variables that can be acquired at minimal cost and effort, and can be applied independent of non-empirical clinical data. In conclusion, the metabolomics profiling data and the modeling work stemming from it, as presented here, could further explain the cause of COVID-19 disease prognosis and patient management.
Nelson Soares 
