Fig. 6: Survival analysis among clusters of critical COVID-19 patients based on trajectories of biomarkers sRAGE and GDF15.

A Survival curves of the three sRAGE-trajectory-based clusters of critical COVID-19 patients (as defined by longitudinal k-means clustering) were plotted using the Kaplan-Meier method. B To assess the effect of the trajectory/kinetics of sRAGE levels in critical COVID-19 patients on the hazard ratio of mortality, Cox proportional hazards modeling was performed with adjustment for age, gender, body mass index and levels of interleukin-6. Patients with sustained moderate sRAGE levels (cluster A) had a 277% higher hazard of mortality compared to patients with a trajectory that started with (and maintained) low sRAGE levels (cluster B), independent from other covariates. C Using the Kaplan-Meier method, survival curves of the three GDF15-trajectory-based clusters of critical COVID-19 patients (as defined by longitudinal k-means clustering) were plotted. D Cox proportional hazards modeling showed that patients with persistent moderate levels of GDF15 on the three consecutive timepoints (cluster A) have a higher hazard of mortality than those in whom the GDF15 start low and remain low (cluster C), taking into account the covariates age, gender, body mass index and interleukin-6 levels.