(891) Endothelial Biomarkers Predict Persistent Pediatric Sepsis-Associated Acute Respiratory Dysfunction

Introduction: Sepsis is a common cause of acute respiratory distress syndrome (ARDS) with mortality rates up to 30%. The heterogeneity of both syndromes has complicated the development of new therapies; thus, current therapy focuses on supportive care. Prognostic enrichment is one strategy to reduce heterogeneity and identify patients at greatest risk of poor outcomes. We hypothesized a model combining clinical variables with biomarkers of endothelial dysfunction could accurately identify critically ill pediatric patients at risk for persistent sepsis associated acute respiratory dysfunction (SA-ARD).

Methods: Secondary analysis of a prospective observational cohort of pediatric septic shock patients. The primary outcome was Day 3 SA-ARD defined as at least one of the following: endotracheal intubation, mechanical ventilation, P/F < 250, PCO2 > 65 torr or PaO2 < 40 torr. Endothelial biomarkers were measured on day 1. Multivariable logistic regression was used to identify biomarkers independently associated with D3 SA-ARD. CART and TreeNet analyses were used to determine which combination of endothelial biomarkers and clinical parameters yielded the model predictive of D3 SA-ARD in the derivation set. Performance of the newly derived model was tested in a hold-out validation set.

Results: Patients with D3 SA-ARD had higher rates of complicated course, 28-day mortality, and ICU length of stay. Soluble thrombomodulin (sTM) was independently associated with D3 SA-ARD (OR 3.24, 95% CI 1.07, 9.80), while VCAM-1 was associated with decreased risk of D3 SA-ARD (OR 0.16, 95% CI 0.06, 0.41). TreeNet and CART models incorporating clinical variables and endothelial biomarkers had an AUROC of 0.96 and 0.88 in the derivation cohort. The CART model which incorporated presence of D1 SA-ARD, P/F ratio < 250, sTM, and VCAM-1 had comparable performance in the hold-out validation set with a sensitivity of 86% and a specificity of 74% for predicting D3 SA-ARD.

Conclusions: We derived and validated a risk prediction model that incorporated endothelial biomarkers to predict patients at high-risk of persistent SA-ARD. Pending prospective validation, our model may be used for prognostic enrichment in targeted trials of therapeutic strategies to mitigate risk associated with pediatric SA-ARD.