Biomarkers for bacteremia and severe sepsis in hematological patients with neutropenic fever: multivariate logistic regression analysis and factor analysis

We compared biomarkers and their changes as predictors for bacteremia and severe sepsis during neutropenic fever after intensive chemotherapy in hematological patients. Serum C-reactive protein (CRP), semi-quantative procalcitonin, aminoterminal pro-brain natriuretic peptide (NT-proBNP), cortisol, lactate, plasma antithrombin and fibrinogen were measured daily from day 0 to day 3/day 4 in 89 neutropenic fever episodes of 65 hematological patients. The best predictors for bacteremia and gram-negative bacteremia were procalcitonin and its change, with odds ratios (ORs) and 95% confidence intervals of 2.63 (1.56-4.44) and 3.20 (1.77-5.80) for bacteremia and 4.14 (2.00-8.58) and 5.04 (2.18-11.63) for gram-negative bacteremia, respectively. For severe sepsis, the best predictors were CRP and fibrinogen, with ORs of 1.94 (1.07-3.52) and 1.92 (1.05-3.54). Factor analysis provided two predictive factors: procalcitonin-NT-proBNP-antithrombin factor predicted gram-negative bacteremia and CRP-fibrinogen predicted severe sepsis. Applying a combination of markers reflecting different aspects of infection might improve the recognition of risk for complications in patients with neutropenic fever.