Abstract
Abstract
Background
The aim of this study was to determine whether the addition of immunologic biosignatures to a predictive model based solely on clinical data could improve the model's ability to predict patient outcome among solid organ transplant (SOT) recipients with gram-negative bacteremia (GNB).
Methods
Plasma was prospectively collected from SOT recipients with GNB. Twenty-seven cytokines, chemokines, and immunoglobulin levels were quantified. Two logistic regression models were specified: a clinical model including age, transplanted organ, and time since transplant and a clinical-immunologic model that additionally incorporated the first two principal components. Model performance was evaluated using area under the curve (AUC). Internal validation was performed by repeated 10-fold cross-validation. AUCs were compared with a paired t test.
Results
Of 169 SOT recipients with GNB, 45 (26.6%) experienced an adverse outcome (septic shock or death). When compared with the clinical model alone, the clinical-immunologic model demonstrated superior discrimination (AUC, 0.70 vs 0.84; P
Conclusions
Immune biosignatures significantly improved prediction of adverse outcomes beyond clinical factors in SOT recipients with GNB.