Evaluation of a risk-guided strategy for empirical carbapenem use in febrile neutropenia.

Febrile neutropenia (FN) is associated with substantial morbidity and necessitates empirical broad-spectrum antimicrobial treatment. In this prospective cohort study, a risk-guided management strategy for FN using empirical piperacillin/tazobactam (TZP) or a carbapenem was evaluated. The analysis involved 723 FN episodes in hospitalised adult patients, including those with severe sepsis or prior infection/colonisation with extended-spectrum β-lactamase (ESBL)-producing Enterobacteriaceae. Propensity score matching analysis was used to adjust for baseline differences between treatment groups and produced 267 matched pairs. The primary outcome was all-cause mortality. Secondary outcomes were the incidences of drug-resistant Gram-negative (including ESBL-producing) and Gram-positive bacterial isolates and of invasive pulmonary aspergillosis (IPA) and their associated mortality. There was no difference in mortality between empirical carbapenem and TZP [18/267 (6.7%) vs. 14/267 (5.2%); P = 0.466]. Higher incidences of drug-resistant Gram-negative isolates [77/267 (28.8%) vs. 26/267 (9.7%); P < 0.001], including ESBL-producing bacteria [57/267 (21.3%) vs. 16/267 (6.0%); P < 0.001], were observed in carbapenem-treated episodes where its use lowered mortality. Mortality rates for ESBL-positive infections were 5.3% (3/57) and 25.0% (4/16) (P = 0.037) and for drug-resistant Gram-negative infections were 6.5% (5/77) and 23.1% (6/26) (P = 0.018) in carbapenem- and TZP-treated episodes, respectively. More IPA was observed with carbapenem use [16/267 (6.0%) vs. 6/267 (2.2%); P = 0.029]. Antifungal prophylaxis reduced the risk of death (odds ratio = 0.39, 95% confidence interval 0.17-0.87; P = 0.017). Risk-guided carbapenem prescribing in FN correctly identified cases prone to drug-resistant Gram-negative infections and reduced the mortality in these episodes.