time–dependent clinical subphenotypes in patients with COVID–19

LD Bos, M Sjoding, P Sinha, SV Bhavani, P Lyons, AF Bewley, M Botta, AM Tsonas, A Serpa Neto,

MJ Schultz, R Dickson, F Paulus, for the PRoVENT–COVID collaborative group.

Introduction

COVID–19 related acute respiratory distress syndrome has been postulated to present with distinct respiratory subphenotypes. Most such phenotyping schema, however, have been limited by sample size, disregard for temporal dynamics and lack of validation. The objective of this study was to identify respiratory subphenotypes of COVID–19 related ARDS using unbiased data–driven approaches.

Methods

Consecutive invasively ventilated patients from 22 ICUs in the Netherlands served as the derivation cohort, and similar patients from two ICUs in the USA as the replication cohort. Latent class analysis (LCA) was used to identify subphenotypes using clinically available respiratory data cross–sectionally at baseline, and longitudinally using 8–hourly data from the first four days of invasive ventilation. Group–based trajectory modeling (GBTM) was used to evaluate trajectories of individual variables and to facilitate potential clinical translation.

Results

1007 patients were included in the derivation cohort, 288 and 326 in the respective replication cohorts. Cross-sectional LCA did not identify any underlying subphenotypes. Longitudinal LCA identified two distinct subphenotypes. Subphenotype 2 was characterized by higher mechanical power (MP), minute ventilation, ventilatory ratio (VR) than Subphenotype 1, while PaO2/FiO2, pH and compliance of the respiratory system were not different. Mortality was not different between the subphenotypes, however, Subphenotype 2 had longer duration of mechanical ventilation and higher venothrombotic events. GBTM revealed trajectories of VR and MP with similar dynamics to those observed in LCA–derived trajectory subphenotypes. However, upward trajectories were better independent prognosticators for 30-day mortality (OR_VR: 1·69, 95%–CI: 1·11–2·57 – OR_MP: 2·28, 95%–CI: 1·50–3·47). The association between upward VR trajectories and 30-day mortality was confirmed in the replication cohorts (OR_{VRvalidation1}: 4·65, 95%–CI: 1·871–1·6; OR_{VRvalidation2}: 1·89, 95%–CI: 1·05–3·37).

Conclusions

At baseline, COVID–19 related ARDS has no consistent respiratory subphenotype. Patients diverge from a a fairly homogenous to a more heterogeneous population with trajectories of ventilatory ratio and mechanical power being the most discriminatory. Modelling these parameters alone provided prognostic value for duration of mechanical ventilation and mortality.