Empirical carbapenems or piperacillin/tazobactam for infections in intensive care: An international retrospective cohort study.

Abstract

Critically ill patients in intensive care units (ICU) are frequently administered broad-spectrum antibiotics (e.g., carbapenems or piperacillin/tazobactam) for suspected or confirmed infections. This retrospective cohort study aimed to describe the use of carbapenems and piperacillin/tazobactam in two international, prospectively collected datasets. We conducted a post hoc analysis of data from the "Adjunctive Glucocorticoid Therapy in Patients with Septic Shock" (ADRENAL) trial (n = 3713) and the "Antimicrobial de-escalation in the critically ill patient and assessment of clinical cure" (DIANA) study (n = 1488). The primary outcome was the proportion of patients receiving initial antibiotic treatment with carbapenems and piperacillin/tazobactam. Secondary outcomes included mortality, days alive and out of ICU and ICU length of stay at 28 days. In the ADRENAL trial, carbapenems were used in 648 out of 3713 (17%), whereas piperacillin/tazobactam was used in 1804 out of 3713 (49%) participants. In the DIANA study, carbapenems were used in 380 out of 1480 (26%), while piperacillin/tazobactam was used in 433 out of 1488 (29%) participants. Mortality at 28 days was 23% for patients receiving carbapenems and 24% for those receiving piperacillin/tazobactam in ADRENAL and 23% and 19%, respectively, in DIANA. We noted variations in secondary outcomes; in DIANA, patients receiving carbapenems had a median of 13 days alive and out of ICU compared with 18 days among those receiving piperacillin/tazobactam. In ADRENAL, the median hospital length of stay was 27 days for patients receiving carbapenems and 21 days for those receiving piperacillin/tazobactam. In this post hoc analysis of ICU patients with infections, we found widespread initial use of carbapenems and piperacillin/tazobactam in international ICUs, with the latter being more frequently used. Randomized clinical trials are needed to assess if the observed variations in outcomes may be drug-related effects or due to confounders.